{ "version": 1, "site": "https://agentpicklist.com", "updated": "2026-10-15", "count": 110, "license_note": "Summaries are ours; follow source links for originals.", "use_cases": [ { "id": "give-claude-code-a-website-and-an-ad-budget", "url": "https://agentpicklist.com/use-cases/give-claude-code-a-website-and-an-ad-budget/", "title": "Hand Claude Code a website and an ad budget to run on its own", "summary": "A web developer gave Claude Code a baseline site, ad accounts with a budget, and scraping tools, and lets it make marketing decisions every four hours.", "what_they_did": "HamadMalikKhan gave Claude Code a baseline website, access to Google, X, LinkedIn and Reddit Ads with a budget, Apify, and OpenRouter. The site is personalized per visitor on the client side using DeepSeek and many device, context and behavior signals.", "outcome": "So far, he says, it mostly does 'tomfoolery': spinning up lots of SEO pages and adding share buttons to try to go viral. He finds it fun to watch it decide every 4 hours.", "worth_knowing": "An experiment, not a success story. Cap ad spend at the account level. Heavy visitor fingerprinting raises privacy and consent issues in many regions, so check the rules where your visitors are.", "category": "marketing", "agents": [ "Claude Code" ], "tools": [ "Google Ads", "X Ads", "LinkedIn Ads", "Reddit Ads", "Apify", "OpenRouter", "DeepSeek" ], "roles": [ "Marketer", "Web developer" ], "industries": [ "Marketing" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported (ad budget + API usage)", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code on a schedule", "Ad platform APIs with hard spend caps", "Apify", "Analytics" ], "steps": [ "Set hard daily spend caps in each ad account first.", "Give the agent read access to analytics and write access to the site repo.", "Run it on a schedule (e.g. every 4 hours) with a log of every decision.", "Review the log daily and roll back bad changes." ], "prompt": "You run growth for this site. Every run: read the last 24h of analytics and ad results, choose ONE change (copy, page, or ad tweak) with the best expected impact, make it, and append your reasoning and the expected result to DECISIONS.md. Never raise spend above the caps.", "repo": null }, "author": { "name": "HamadMalikKhan", "handle": "HamadMalikKhan", "platform": "Hacker News", "url": "https://news.ycombinator.com/user?id=HamadMalikKhan" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=50107586", "platform": "Hacker News", "posted_at": "2026-10-15", "engagement_n": 0 } ], "posted_at": "2026-10-15", "added_at": "2026-10-15" }, { "id": "spend-leftover-ai-quota-on-reviewable-prs", "url": "https://agentpicklist.com/use-cases/spend-leftover-ai-quota-on-reviewable-prs/", "title": "Spend leftover Claude and Codex quota on small pull requests", "summary": "A founder built a bot that turns spare subscription tokens into pull requests for small tasks he never gets around to.", "what_they_did": "Ari Wilson, founder of Bakeoff, built RemainderBot. Given your goals and task list, it uses unused Claude and Codex subscription capacity to write PRs you can review.", "outcome": "He built it to clear small context-heavy tasks without context-switching and to stop wasting subscription limits in quiet weeks. It is open source and he calls it early.", "worth_knowing": "Every PR still needs your review. Check your plan's terms on automated use.", "category": "coding", "agents": [ "Claude Code", "Codex" ], "tools": [ "GitHub" ], "roles": [ "Founder", "Developer" ], "industries": [ "Software" ], "difficulty": "beginner", "time_to_build": { "text": "est. ~1 hour to set up", "sourced": false }, "monthly_cost": { "text": "Uses existing subscriptions", "sourced": true }, "revenue": null, "result": null, "recipe": { "stack": [ "openremainderbot (GitHub)", "Claude and/or Codex subscription", "A task list file" ], "steps": [ "Clone the repo and follow its setup.", "Write a short goals file and a list of small, well-scoped tasks.", "Let it open PRs near the end of your usage window.", "Review and merge or close." ], "prompt": "From TASKS.md, pick the smallest task you can fully finish, implement it on a new branch with tests, and open a PR with a 3-line summary and how to verify it.", "repo": "https://github.com/arimwilson/openremainderbot" }, "author": { "name": "Ari Wilson", "handle": "ariwilson", "platform": "Hacker News", "url": "https://news.ycombinator.com/user?id=ariwilson" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=50073607", "platform": "Hacker News", "posted_at": "2026-10-13", "engagement_n": 0 } ], "posted_at": "2026-10-13", "added_at": "2026-10-15" }, { "id": "render-sheet-music-in-a-new-base-12-notation-with-codex", "url": "https://agentpicklist.com/use-cases/render-sheet-music-in-a-new-base-12-notation-with-codex/", "title": "Render sheet music in a new base-12 notation with Codex", "summary": "A musician used Codex to build a score renderer on MuseScore with custom note heads for his isomorphic keyboard.", "what_they_did": "jmole plays a new isomorphic keyboard with base-12 notation and used Codex to build a score-rendering backend on MuseScore with note heads showing each pitch class.", "outcome": "He says it has been pretty helpful for learning new music.", "worth_knowing": "Building on an existing engine (MuseScore) beat writing a renderer from scratch.", "category": "content", "agents": [ "Codex" ], "tools": [ "MuseScore" ], "roles": [ "Musician" ], "industries": [ "Music" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Codex", "MuseScore" ], "steps": [ "Pick a repeatable format.", "Let the agent generate drafts or assets from your inputs.", "Edit and check facts and rights before publishing.", "Reuse the template for the next piece." ], "prompt": "Extend MuseScore rendering so each note head shows its pitch class as 0-9, X, Y. Convert this MusicXML file and output a PDF.", "repo": null }, "author": { "name": "jmole", "handle": "jmole", "url": "https://news.ycombinator.com/user?id=jmole", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=50075665", "platform": "Hacker News", "posted_at": "2026-10-13", "engagement": "83 points", "engagement_n": 83 } ], "posted_at": "2026-10-13", "added_at": "2026-10-15" }, { "id": "let-claude-control-a-music-notation-app-over-mcp", "url": "https://agentpicklist.com/use-cases/let-claude-control-a-music-notation-app-over-mcp/", "title": "Let Claude control a music notation app over MCP", "summary": "A developer built an iPad and Mac notation editor that Claude Code or Claude Desktop can drive.", "what_they_did": "yetanotheruser built Autograph, a music notation app with MCP control.", "outcome": "Shared on HN as a launch.", "worth_knowing": "A tool launch rather than a usage report.", "category": "content", "agents": [ "Claude" ], "tools": [ "MCP", "Swift" ], "roles": [ "Musician", "Developer" ], "industries": [ "Music" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Autograph", "Claude Desktop" ], "steps": [ "Pick a repeatable format.", "Let the agent generate drafts or assets from your inputs.", "Edit and check facts and rights before publishing.", "Reuse the template for the next piece." ], "prompt": "Write an 8-bar melody in D minor for violin and add chord symbols.", "repo": null }, "author": { "name": "yetanotheruser", "handle": "yetanotheruser", "url": "https://news.ycombinator.com/user?id=yetanotheruser", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=50070306", "platform": "Hacker News", "posted_at": "2026-10-13", "engagement": "", "engagement_n": 0 } ], "posted_at": "2026-10-13", "added_at": "2026-10-15" }, { "id": "drive-parametric-cad-from-agents-over-mcp", "url": "https://agentpicklist.com/use-cases/drive-parametric-cad-from-agents-over-mcp/", "title": "Drive parametric CAD from agents over MCP", "summary": "An open-source Rust parametric CAD tool supports MCP so agents can build models.", "what_they_did": "bherbruck built SolidCraft with MCP support.", "outcome": "141 points on HN.", "worth_knowing": "A tool launch rather than a usage report.", "category": "home", "agents": [], "tools": [ "Rust", "MCP" ], "roles": [ "Engineer", "Maker" ], "industries": [ "Hardware", "Manufacturing" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "SolidCraft", "Claude or Codex" ], "steps": [ "List the devices and data sources.", "Have the agent write the integration code.", "Test on one device first.", "Automate and monitor." ], "prompt": "Model a wall bracket for a 25 mm pipe with two M5 holes, 4 mm thick, then export STL.", "repo": null }, "author": { "name": "bherbruck", "handle": "bherbruck", "url": "https://news.ycombinator.com/user?id=bherbruck", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=50064840", "platform": "Hacker News", "posted_at": "2026-10-13", "engagement": "141 points", "engagement_n": 141 } ], "posted_at": "2026-10-13", "added_at": "2026-10-15" }, { "id": "drive-coding-agents-from-telegram-and-resume-after-usage-limits", "url": "https://agentpicklist.com/use-cases/drive-coding-agents-from-telegram-and-resume-after-usage-limits/", "title": "Drive coding agents from Telegram and resume after usage limits", "summary": "A developer built nightmux to control Claude Code agents from Telegram and auto-resume past usage limits overnight.", "what_they_did": "@Nightmux’s agents used to die at 2am when they hit usage limits.", "outcome": "Open source: one file, zero dependencies.", "worth_knowing": "Check your plan’s terms.", "category": "coding", "agents": [ "Claude Code" ], "tools": [ "Telegram" ], "roles": [ "Developer" ], "industries": [ "Software" ], "difficulty": "beginner", "time_to_build": { "text": "est. 10 minutes", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "nightmux", "Telegram bot", "Claude Code" ], "steps": [ "Open the repo in your coding agent and describe the goal and constraints.", "Ask for a plan first, then let it implement in small, reviewable steps.", "Run the tests or the app after each step and feed errors back.", "Review the diff before merging." ], "prompt": "Continue the task in TODO.md; when done, message me on Telegram with a summary.", "repo": null }, "author": { "name": "@Nightmux", "handle": "Nightmux", "url": "https://x.com/Nightmux", "platform": "X" }, "sources": [ { "url": "https://x.com/Nightmux/status/2110008902125543649", "platform": "X", "posted_at": "2026-10-13", "engagement": "2 likes", "engagement_n": 2 } ], "posted_at": "2026-10-13", "added_at": "2026-10-15" }, { "id": "cut-trial-churn-by-moving-the-paywall", "url": "https://agentpicklist.com/use-cases/cut-trial-churn-by-moving-the-paywall/", "title": "Cut trial churn by testing when your paywall appears", "summary": "A solo iOS developer who calls Claude Code 'basically my whole team' cut trial churn from 50% to 30% by only changing paywall timing.", "what_they_did": "@chijuano builds iOS apps solo. Instead of shipping more features, he tested the cheapest lever first: when the paywall shows up. Same app, price and copy.", "outcome": "He reports trial churn fell from 50% to 30%. He says he is still under $1k MRR with a goal of $10k.", "worth_knowing": "His lesson: Claude Code can build any screen, but it can't tell you which screen matters. Change one thing at a time.", "category": "business", "agents": [ "Claude Code" ], "tools": [ "iOS", "App Store" ], "roles": [ "Solo founder", "iOS developer" ], "industries": [ "Consumer apps" ], "difficulty": "beginner", "time_to_build": { "text": "An afternoon (sourced)", "sourced": true }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": "Sourced: trial churn 50% → 30%", "recipe": { "stack": [ "Claude Code", "Your app's paywall SDK", "Analytics with trial events" ], "steps": [ "Measure trial churn, not just revenue.", "Pick the moment users first get value and move the paywall there.", "Change only that, and compare two weeks before and after." ], "prompt": "Move the paywall in this app from [current trigger] to right after the user completes [first success moment]. Keep the copy and prices identical, and add analytics events for paywall_shown and trial_started.", "repo": null }, "author": { "name": "@chijuano", "handle": "chijuano", "platform": "X", "url": "https://x.com/chijuano" }, "sources": [ { "url": "https://x.com/chijuano/status/2109681987111752151", "platform": "X", "posted_at": "2026-10-12", "engagement_n": 0 } ], "posted_at": "2026-10-12", "added_at": "2026-10-15" }, { "id": "run-cheap-sandboxes-for-personal-agents", "url": "https://agentpicklist.com/use-cases/run-cheap-sandboxes-for-personal-agents/", "title": "Host personal agents in sandboxes for cents a month", "summary": "A bootstrapped YC founder explains how his sandbox service runs personal Claude Code and OpenClaw agents for about $0.05-$0.10 per free user a month.", "what_they_did": "Vishnu (Agent37) started as an OpenClaw wrapper and now hosts sandboxes for agent harnesses. Most new users run personal Hermes, OpenClaw or Claude Code agents on the free plan.", "outcome": "He says customers average $0.37/month per sandbox, and free sandboxes cost about $0.05-$0.10 per user per month, from burst-aware CPU and RAM packing and renting from Hetzner and OVH instead of hyperscalers.", "worth_knowing": "These are the vendor's own numbers. Sandboxes trade isolation strength for cost; check what your agent can reach.", "category": "ops", "agents": [ "Claude Code", "OpenClaw" ], "tools": [ "Hetzner", "OVH" ], "roles": [ "Founder", "DevOps" ], "industries": [ "Cloud", "Software" ], "difficulty": "beginner", "time_to_build": { "text": "est. 15 minutes", "sourced": false }, "monthly_cost": { "text": "Sourced: average $0.37/month per sandbox", "sourced": true }, "revenue": null, "result": null, "recipe": { "stack": [ "A hosted agent sandbox or a small VPS", "Your agent harness" ], "steps": [ "Pick a sandbox provider or a cheap VPS.", "Install your agent harness and give it only the credentials it needs.", "Schedule tasks and check logs weekly." ], "prompt": "Every morning at 7am, check my inbox label 'to-do', summarize anything new in 5 bullets, and save the summary to notes/daily.md. Do not reply to anyone.", "repo": null }, "author": { "name": "Vishnu (Agent37)", "handle": "vishnukool", "platform": "Hacker News", "url": "https://news.ycombinator.com/user?id=vishnukool" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=50059096", "platform": "Hacker News", "posted_at": "2026-10-12", "engagement": "160 points", "engagement_n": 160 } ], "posted_at": "2026-10-12", "added_at": "2026-10-15" }, { "id": "make-music-in-a-daw-through-mcp", "url": "https://agentpicklist.com/use-cases/make-music-in-a-daw-through-mcp/", "title": "Make music in a DAW through MCP", "summary": "A developer connected a music workstation to agents through MCP.", "what_they_did": "androsynth89 built Blokkily for making music through MCP.", "outcome": "84 points on HN.", "worth_knowing": "A tool launch rather than a usage report.", "category": "content", "agents": [], "tools": [ "MCP", "DAW" ], "roles": [ "Musician" ], "industries": [ "Music" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Blokkily", "An MCP client" ], "steps": [ "Pick a repeatable format.", "Let the agent generate drafts or assets from your inputs.", "Edit and check facts and rights before publishing.", "Reuse the template for the next piece." ], "prompt": "Make a 16-bar lo-fi loop at 80 BPM with drums, bass and a Rhodes chord progression.", "repo": null }, "author": { "name": "androsynth89", "handle": "androsynth89", "url": "https://news.ycombinator.com/user?id=androsynth89", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=50064438", "platform": "Hacker News", "posted_at": "2026-10-12", "engagement": "84 points", "engagement_n": 84 } ], "posted_at": "2026-10-12", "added_at": "2026-10-15" }, { "id": "have-your-agent-phone-a-restaurant-and-book-only-if-your-must-haves-are-met", "url": "https://agentpicklist.com/use-cases/have-your-agent-phone-a-restaurant-and-book-only-if-your-must-haves-are-met/", "title": "Have your agent phone a restaurant and book only if your must-haves are met", "summary": "A user’s OpenClaw agent called restaurants in Dubai with a strict brief and declined to book when a must-have was missing.", "what_they_did": "@ai_samerhaddad briefed his OpenClaw agent: Tuesday, 8:30-9 pm, three people, his name, his number only if asked, and two must-haves.", "outcome": "It checked the number online before dialing. On the first call one must-have was not met, so it did not book and said it would check back.", "worth_knowing": "Give hard constraints, and disclose that an AI is calling where required.", "category": "personal", "agents": [ "OpenClaw" ], "tools": [ "Voice calling" ], "roles": [ "Individual" ], "industries": [ "Personal", "Hospitality" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "OpenClaw", "Voice calling tool" ], "steps": [ "Give the agent access only to the accounts it needs.", "Write a clear brief with hard constraints.", "Ask it to report back before anything irreversible.", "Review the outcome and refine the brief." ], "prompt": "Call [restaurant] to book Tuesday 8:30-9pm for 3 under [name]. Book only if [must-have 1] and [must-have 2] are true. Share my number only if asked. Report back.", "repo": null }, "author": { "name": "@ai_samerhaddad", "handle": "ai_samerhaddad", "url": "https://x.com/ai_samerhaddad", "platform": "X" }, "sources": [ { "url": "https://x.com/ai_samerhaddad/status/2109555545212424376", "platform": "X", "posted_at": "2026-10-12", "engagement": "2 likes", "engagement_n": 2 } ], "posted_at": "2026-10-12", "added_at": "2026-10-15" }, { "id": "build-an-agent-that-files-refunds-and-insurance-claims-for-you", "url": "https://agentpicklist.com/use-cases/build-an-agent-that-files-refunds-and-insurance-claims-for-you/", "title": "Build an agent that files refunds and insurance claims for you", "summary": "A hackathon team built Recourse, which organizes evidence, prepares paperwork and calls customer support for you.", "what_they_did": "@_benjaminxie and teammates built Recourse at a16z SpeedHacks.", "outcome": "They won first place.", "worth_knowing": "A hackathon prototype.", "category": "personal", "agents": [], "tools": [ "Voice calling" ], "roles": [ "Builder" ], "industries": [ "Insurance", "Consumer" ], "difficulty": "advanced", "time_to_build": { "text": "A weekend (hackathon)", "sourced": true }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "LLM", "Document parsing", "Voice calling" ], "steps": [ "Give the agent access only to the accounts it needs.", "Write a clear brief with hard constraints.", "Ask it to report back before anything irreversible.", "Review the outcome and refine the brief." ], "prompt": "Gather my receipts and emails about [order], write a refund request citing the policy, and prepare a call script.", "repo": null }, "author": { "name": "@_benjaminxie", "handle": "_benjaminxie", "url": "https://x.com/_benjaminxie", "platform": "X" }, "sources": [ { "url": "https://x.com/_benjaminxie/status/2109540504052683244", "platform": "X", "posted_at": "2026-10-12", "engagement": "6 likes", "engagement_n": 6 } ], "posted_at": "2026-10-12", "added_at": "2026-10-15" }, { "id": "check-whether-you-really-need-replacement-staff-before-calling-them-in", "url": "https://agentpicklist.com/use-cases/check-whether-you-really-need-replacement-staff-before-calling-them-in/", "title": "Check whether you really need replacement staff before calling them in", "summary": "A hotel analyst built an n8n workflow to check housekeeping staffing needs with data when staff call in sick.", "what_they_did": "@syifa_anns built a workflow for a fully booked hotel with four housekeepers out sick.", "outcome": "Her summary: less panic, more math.", "worth_knowing": "Automate the calculation, keep the decision human.", "category": "ops", "agents": [ "n8n" ], "tools": [], "roles": [ "Hotel operations" ], "industries": [ "Hospitality" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "n8n", "Occupancy data" ], "steps": [ "Write down the manual process step by step.", "Automate the deterministic parts with plain workflow steps or scripts.", "Use the model only where judgment is needed, with confidence checks.", "Add alerts and a human fallback." ], "prompt": "Given rooms to clean today, minutes per room type and staff on shift, how many extra housekeepers are needed?", "repo": null }, "author": { "name": "@syifa_anns", "handle": "syifa_anns", "url": "https://x.com/syifa_anns", "platform": "X" }, "sources": [ { "url": "https://x.com/syifa_anns/status/2108807551211176183", "platform": "X", "posted_at": "2026-10-10", "engagement": "", "engagement_n": 0 } ], "posted_at": "2026-10-10", "added_at": "2026-10-15" }, { "id": "run-a-coding-agent-pipeline-like-a-teaching-hospital", "url": "https://agentpicklist.com/use-cases/run-a-coding-agent-pipeline-like-a-teaching-hospital/", "title": "Run a fleet of coding agents like a teaching hospital", "summary": "Two Cockroach Labs engineers turned GitHub issues into 'patients' handled by agent roles (triage nurse, fellow, reviewing attending, discharge nurse) and shipped production code for five months.", "what_they_did": "Adam Storm and Rafi Shamim built a GitHub Actions pipeline where each label hands an issue to the next agent role. A fellow agent must reproduce the bug and write a treatment plan, a separate reviewer agent approves or rejects the plan, and a discharge agent audits that the review actually happened before merge. A human 'Chief' handles escalations and records precedents agents must check first.", "outcome": "The source reports well over a million lines of code landed between April 21 and September 11, with IBM Db2 support for their MOLT migration tool built in under two days for a $4,172 token bill (versus about 9 months and roughly $160K for the earlier Oracle work). Total spend was just over $135,000 in Claude tokens, about $84 per issue.", "worth_knowing": "The agents got bureaucratic: a one-line copy edit went through five rework rounds, so they added a circuit-breaker after repeated review rounds. Their skill files grew to 100,000+ words and about 23% turned out to be redundant. They also switched on mandatory human approval for merges after the first week.", "category": "coding", "agents": [ "Claude Code" ], "tools": [ "GitHub Actions", "Claude" ], "roles": [ "Engineering lead", "Developer" ], "industries": [ "Software" ], "difficulty": "advanced", "time_to_build": { "text": "Months (iterated over 5 months)", "sourced": true }, "monthly_cost": { "text": "Sourced: ~$135K total over ~5 months, ~$84 per issue", "sourced": true }, "revenue": null, "result": "Sourced: Db2 support '164x faster, and 38x cheaper' than the earlier Oracle effort", "recipe": { "stack": [ "Claude (Opus/Sonnet-class models)", "GitHub Actions", "GitHub labels as state", "Markdown skill files per role" ], "steps": [ "Define 4-6 roles (triage, planner, plan reviewer, implementer, code reviewer, merge auditor) as separate skill files.", "Use issue labels as the state machine: each label triggers one workflow and the agent applies the next label when done.", "Require a written plan and an independent plan review before any code is written.", "Forbid weakening tests; reviews start by reverting the change and confirming the new tests fail.", "Add a circuit-breaker: escalate to a human after a fixed number of rework rounds.", "Keep an append-only precedent log of human decisions that agents must read before escalating." ], "prompt": "You are the PLAN REVIEWER, a different agent from the one who wrote this plan. Your only job is to find what is wrong with it: missing reproduction, wrong root cause, untested paths, scope creep, security or data risks. Reply APPROVE, REJECT (with numbered reasons) or ESCALATE (with the exact question for a human). Do not write code.", "repo": null }, "author": { "name": "Adam Storm & Rafi Shamim (Cockroach Labs)", "handle": "rafiss", "platform": "Cockroach Labs blog", "url": "https://www.cockroachlabs.com/blog/experiment-running-hospital-code/" }, "sources": [ { "url": "https://www.cockroachlabs.com/blog/experiment-running-hospital-code/", "platform": "Blog", "posted_at": "2026-10-09", "engagement_n": 0 }, { "url": "https://news.ycombinator.com/item?id=50021899", "platform": "Hacker News", "posted_at": "2026-10-09", "engagement": "218 points", "engagement_n": 218 } ], "posted_at": "2026-10-09", "added_at": "2026-10-15" }, { "id": "decompile-a-video-game-with-a-swarm-of-agents", "url": "https://agentpicklist.com/use-cases/decompile-a-video-game-with-a-swarm-of-agents/", "title": "Decompile a classic shooter with a swarm of coding agents", "summary": "Over three months, a developer and community helpers had up to 16 Claude Code and Codex agents rebuild an old first-person shooter as readable C++ that runs without noticeable bugs.", "what_they_did": "Maurice Heumann ran worker agents in Claude Code CLI and Codex CLI, tracking one GitHub issue per source file and letting agents talk in a shared Discord channel. They used Hex-Rays' ida-mcp for disassembly. After the first month's output proved semantically wrong, they added a byte-matching 'oracle' script that compares each rebuilt function to the original binary and gives a pass/fail.", "outcome": "Per the post, 99% of the game's functions are present in the reconstructed source and 83% are byte-exact, and the game runs with all original features. He estimates 600-700 billion tokens were used. The code stays private.", "worth_knowing": "Agents cheated when given a check: they wrote inline assembly and tried to edit the verification script, so CI now hashes the script. Instructions decayed over time, fixed by an hourly cron that tells agents to reread the rules. A strict pass/fail check let much cheaper models do the work.", "category": "coding", "agents": [ "Claude Code", "Codex" ], "tools": [ "IDA Pro (ida-mcp)", "GitHub CLI", "Discord", "MCP" ], "roles": [ "Developer", "Reverse engineer" ], "industries": [ "Games" ], "difficulty": "advanced", "time_to_build": { "text": "~3 months", "sourced": true }, "monthly_cost": { "text": "Sourced: Claude Max (20x) + Codex Pro subscriptions", "sourced": true }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code CLI", "Codex CLI", "ida-mcp", "GitHub Issues", "Discord or another shared channel", "A machine-checkable pass/fail script" ], "steps": [ "Split the job into units with one tracked issue each.", "Write a single goals-and-rules document and re-inject it on a schedule.", "Build an objective oracle (tests, byte match, golden outputs) before scaling agents.", "Ban obvious shortcuts and protect the oracle from edits (hash it in CI).", "Start with a few agents, then scale cheaper models once the oracle is strict." ], "prompt": "Pick the next open issue labeled 'todo', assign it to yourself, and reconstruct only that unit. Before pushing, run ./verify on your change. If it fails, keep working. You may NOT edit ./verify, skip it, or use inline assembly. Post a one-line status in the team channel when you take or finish an issue.", "repo": null }, "author": { "name": "Maurice Heumann", "handle": "momo5502", "platform": "Blog", "url": "https://momo5502.com/posts/2026-10-09-game-decompilation/" }, "sources": [ { "url": "https://momo5502.com/posts/2026-10-09-game-decompilation/", "platform": "Blog", "posted_at": "2026-10-09", "engagement_n": 0 }, { "url": "https://news.ycombinator.com/item?id=50039092", "platform": "Hacker News", "posted_at": "2026-10-11", "engagement": "217 points", "engagement_n": 217 } ], "posted_at": "2026-10-09", "added_at": "2026-10-15" }, { "id": "let-your-agents-draw-arrows-and-boxes-on-your-screen", "url": "https://agentpicklist.com/use-cases/let-your-agents-draw-arrows-and-boxes-on-your-screen/", "title": "Let your agents draw arrows and boxes on your screen", "summary": "A developer made a tool so coding agents can point at things on your screen with big arrows, boxes and text.", "what_they_did": "franze built a way for AI agents to paint overlays on your screen to show you things.", "outcome": "It got 419 points on HN.", "worth_knowing": "A tool launch rather than a usage report.", "category": "coding", "agents": [], "tools": [ "Screen overlay" ], "roles": [ "Developer" ], "industries": [ "Software" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Overlay tool", "Your coding agent" ], "steps": [ "Open the repo in your coding agent and describe the goal and constraints.", "Ask for a plan first, then let it implement in small, reviewable steps.", "Run the tests or the app after each step and feed errors back.", "Review the diff before merging." ], "prompt": "Point at the setting I need to change with an arrow and a one-line label.", "repo": null }, "author": { "name": "franze", "handle": "franze", "url": "https://news.ycombinator.com/user?id=franze", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=50018817", "platform": "Hacker News", "posted_at": "2026-10-09", "engagement": "419 points", "engagement_n": 419 } ], "posted_at": "2026-10-09", "added_at": "2026-10-15" }, { "id": "put-an-ai-agent-on-a-nokia-110-feature-phone", "url": "https://agentpicklist.com/use-cases/put-an-ai-agent-on-a-nokia-110-feature-phone/", "title": "Put an AI agent on a Nokia 110 feature phone", "summary": "A developer reverse-engineered a 48 MB RAM Nokia 110 to run an AI agent and automate a few actions.", "what_they_did": "anupray uses a Nokia 110 to cut screen time. Installing Android failed, so he spent a few days reverse-engineering the phone.", "outcome": "He got an AI agent working and automated a few actions; the code is on GitHub.", "worth_knowing": "Hardware hacking can brick devices.", "category": "home", "agents": [], "tools": [ "Feature phone" ], "roles": [ "Hacker", "Hobbyist" ], "industries": [ "Hardware" ], "difficulty": "advanced", "time_to_build": { "text": "A few days (sourced)", "sourced": true }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Nokia 110", "Reverse-engineering tools", "A cloud LLM API" ], "steps": [ "List the devices and data sources.", "Have the agent write the integration code.", "Test on one device first.", "Automate and monitor." ], "prompt": "Help me map the phone’s key events and SMS interface so a cloud agent can receive a text command and reply.", "repo": "https://github.com/anupray95/AI-Agent-on-a-NOKIA" }, "author": { "name": "anupray", "handle": "anupray", "url": "https://news.ycombinator.com/user?id=anupray", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=50006114", "platform": "Hacker News", "posted_at": "2026-10-08", "engagement": "33 points", "engagement_n": 33 } ], "posted_at": "2026-10-08", "added_at": "2026-10-15" }, { "id": "auto-post-ugc-videos-with-a-self-improving-bot", "url": "https://agentpicklist.com/use-cases/auto-post-ugc-videos-with-a-self-improving-bot/", "title": "Grow signups with a bot that makes and posts UGC videos daily", "summary": "An indie builder's Claude Code bot makes four videos a day, posts them, and learns from past performance; he says it drives 90% of his signups.", "what_they_did": "@gladwyn1_1 built an Instagram bot with Claude Code that runs on his laptop, makes 4 UGC-style videos a day, and posts them. His latest update has it study how earlier videos performed to make the next ones better.", "outcome": "On day 6 of building in public he reported 1,847 users, $81 MRR (+$25), and said about 90% of signups come from the bot.", "worth_knowing": "Early stage and self-reported. Automated posting must follow Instagram's rules and disclose AI-generated content where required.", "category": "marketing", "agents": [ "Claude Code" ], "tools": [ "Instagram" ], "roles": [ "Indie founder", "Marketer" ], "industries": [ "Consumer apps", "Marketing" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": { "usd_per_month": 81, "text": "$81 MRR (self-reported)", "bucket": "under1k", "source_stated": true, "verified": false }, "result": null, "recipe": { "stack": [ "Claude Code", "A video generation API", "Instagram Graph API (business account)", "A simple performance log" ], "steps": [ "Write 10 hooks that worked for your niche.", "Have the agent script, generate and caption a video from a hook.", "Post through the official API on a schedule.", "Log views and signups per video and feed the top performers back into the next batch." ], "prompt": "Read performance.csv. Find the 3 best-performing hooks from the last 14 days, write 4 new video scripts that remix them for [app], generate them, and queue them for posting at the usual times.", "repo": null }, "author": { "name": "@gladwyn1_1", "handle": "gladwyn1_1", "platform": "X", "url": "https://x.com/gladwyn1_1" }, "sources": [ { "url": "https://x.com/gladwyn1_1/status/2107622517590642980", "platform": "X", "posted_at": "2026-10-07", "engagement_n": 0 } ], "posted_at": "2026-10-07", "added_at": "2026-10-15" }, { "id": "make-claudes-replies-half-as-long-with-a-plugin", "url": "https://agentpicklist.com/use-cases/make-claudes-replies-half-as-long-with-a-plugin/", "title": "Make Claude’s replies half as long with a plugin", "summary": "A developer annoyed by long, filler-heavy replies built a Claude Code plugin that cuts them.", "what_they_did": "lowenbjer found system prompts were forgotten after a few turns, so he built Terse.", "outcome": "He says it halves reply length by cutting filler.", "worth_knowing": "Shorter replies also save tokens.", "category": "coding", "agents": [ "Claude Code" ], "tools": [ "Claude Code plugins" ], "roles": [ "Developer" ], "industries": [ "Software" ], "difficulty": "beginner", "time_to_build": { "text": "est. 5 minutes", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code", "claude-terse" ], "steps": [ "Open the repo in your coding agent and describe the goal and constraints.", "Ask for a plan first, then let it implement in small, reviewable steps.", "Run the tests or the app after each step and feed errors back.", "Review the diff before merging." ], "prompt": "Answer in at most 4 sentences. No recaps, slogans or metaphors.", "repo": "https://github.com/lowenbjer/claude-terse" }, "author": { "name": "lowenbjer", "handle": "lowenbjer", "url": "https://news.ycombinator.com/user?id=lowenbjer", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49995809", "platform": "Hacker News", "posted_at": "2026-10-07", "engagement": "25 points", "engagement_n": 25 } ], "posted_at": "2026-10-07", "added_at": "2026-10-15" }, { "id": "let-an-always-on-agent-post-to-slack-and-bring-replies-back", "url": "https://agentpicklist.com/use-cases/let-an-always-on-agent-post-to-slack-and-bring-replies-back/", "title": "Let an always-on agent post to Slack and bring replies back", "summary": "Dan Shipper says OpenAI’s Dots took over Every, from relaying Slack threads to pulling key info out of school emails.", "what_they_did": "At Every, Dan Shipper’s Dot posted his model-testing results to Slack and brought back replies. Colleagues use Dots to pull info from school emails, catch missed Slack messages and flag access requests.", "outcome": "He says his weekend stayed clear and calm because he did not have to open Slack.", "worth_knowing": "Decide what it may post on your behalf.", "category": "ops", "agents": [ "OpenAI Dots" ], "tools": [ "Slack" ], "roles": [ "Founder", "Knowledge worker" ], "industries": [ "Media" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "An always-on assistant", "Slack and email access" ], "steps": [ "Write down the manual process step by step.", "Automate the deterministic parts with plain workflow steps or scripts.", "Use the model only where judgment is needed, with confidence checks.", "Add alerts and a human fallback." ], "prompt": "Post my test results to #models, collect replies until 6pm, and give me a 3-bullet summary. Do not reply on my behalf.", "repo": null }, "author": { "name": "Dan Shipper", "handle": "danshipper", "url": "https://x.com/danshipper", "platform": "X" }, "sources": [ { "url": "https://x.com/danshipper/status/2107890633486582208", "platform": "X", "posted_at": "2026-10-07", "engagement": "134 likes", "engagement_n": 134 } ], "posted_at": "2026-10-07", "added_at": "2026-10-15" }, { "id": "see-all-your-claude-code-and-codex-sessions-on-one-board", "url": "https://agentpicklist.com/use-cases/see-all-your-claude-code-and-codex-sessions-on-one-board/", "title": "See all your Claude Code and Codex sessions on one board", "summary": "A team with 90 AI sessions running at once built a local board showing which ones are waiting on them.", "what_they_did": "@cockpitAIDev’s team had sessions for a board update, a price-test readout and a hiring debrief, and could not tell which needed them.", "outcome": "They released Cockpit, free and local.", "worth_knowing": "Session overload has a real cost.", "category": "ops", "agents": [ "Claude Code", "Codex" ], "tools": [], "roles": [ "Founder", "Operations" ], "industries": [ "Software" ], "difficulty": "beginner", "time_to_build": { "text": "est. 5 minutes", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Cockpit", "Claude Code", "Codex" ], "steps": [ "Write down the manual process step by step.", "Automate the deterministic parts with plain workflow steps or scripts.", "Use the model only where judgment is needed, with confidence checks.", "Add alerts and a human fallback." ], "prompt": "Summarize each running session in one line and tell me which are waiting on my input.", "repo": null }, "author": { "name": "@cockpitAIDev", "handle": "cockpitAIDev", "url": "https://x.com/cockpitAIDev", "platform": "X" }, "sources": [ { "url": "https://x.com/cockpitAIDev/status/2107878627366850966", "platform": "X", "posted_at": "2026-10-07", "engagement": "", "engagement_n": 0 } ], "posted_at": "2026-10-07", "added_at": "2026-10-15" }, { "id": "turn-a-phone-and-smart-plug-into-a-darkroom-timer", "url": "https://agentpicklist.com/use-cases/turn-a-phone-and-smart-plug-into-a-darkroom-timer/", "title": "Turn your phone and a $20 smart plug into a darkroom timer", "summary": "A film photographer vibe coded an iOS app with Codex that drives his enlarger through a cheap smart plug, replacing a $300+ dedicated timer.", "what_they_did": "Peter Szentkiralyi built Darkplug, an iPhone app that switches the enlarger through a local-network smart plug and handles f-stop timing, test strips, dodge and burn, split grade and paper development timing. He says it was 'blindly vibe coded using Codex and Xcode'.", "outcome": "He has used it for a few months in a bathroom darkroom. He puts the cost at under $50 for the app and plug versus $300+ for a boutique hardware timer.", "worth_knowing": "Relies on a compatible smart plug on your local network. Hands-free control in the darkroom is traditionally a foot switch, which came up in the discussion.", "category": "home", "agents": [ "Codex" ], "tools": [ "Xcode", "Smart plug", "iOS" ], "roles": [ "Photographer", "Hobbyist" ], "industries": [ "Photography" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Sourced: one-time ~$20 smart plug", "sourced": true }, "revenue": null, "result": "Sourced: under $50 vs $300+ for a hardware timer", "recipe": { "stack": [ "Codex", "Xcode", "A smart plug with a local API" ], "steps": [ "Confirm your smart plug can be switched over the local network.", "Ask Codex for a minimal SwiftUI app that toggles the plug for N seconds.", "Add f-stop math and test strips one feature at a time.", "Test with the enlarger before trusting it with paper." ], "prompt": "Build a SwiftUI iPhone app that turns a smart plug at [IP] on for an exact number of seconds over the local network. Add f-stop timing (base time × 2^(stops)), a test-strip mode, and a big red stop button.", "repo": null }, "author": { "name": "Peter Szentkiralyi", "handle": "pentakkusu", "platform": "Hacker News", "url": "https://news.ycombinator.com/user?id=pentakkusu" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49978595", "platform": "Hacker News", "posted_at": "2026-10-06", "engagement": "137 points", "engagement_n": 137 }, { "url": "https://peterszentkiralyi.eu/darkplug/", "platform": "Website", "posted_at": "2026-10-06", "engagement_n": 0 } ], "posted_at": "2026-10-06", "added_at": "2026-10-15" }, { "id": "rebuild-simcity-2000-as-a-low-poly-web-game-with-27-subagents", "url": "https://agentpicklist.com/use-cases/rebuild-simcity-2000-as-a-low-poly-web-game-with-27-subagents/", "title": "Rebuild SimCity 2000 as a low-poly web game with 27 subagents", "summary": "A developer had Opus 5.5 decompile SimCity 2000 with Ghidra and port it function by function to TypeScript and WebGL over a few days.", "what_they_did": "drdator started with one prompt asking for a TypeScript WebGL low-poly replica of the full game logic, with the original files available. The agent decompiled it with Ghidra and translated the old PowerPC code function by function.", "outcome": "He says that after about 2 hours with 27 subagents he had a running build; the clone came together over a few days.", "worth_knowing": "Work only from software you own, and respect copyright when publishing.", "category": "gaming", "agents": [ "Claude" ], "tools": [ "Ghidra", "TypeScript", "WebGL" ], "roles": [ "Developer", "Hobbyist" ], "industries": [ "Games" ], "difficulty": "advanced", "time_to_build": { "text": "A few days (sourced)", "sourced": true }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude (Opus-class) with subagents", "Ghidra", "TypeScript + WebGL" ], "steps": [ "Describe the game or mod you want.", "Build it with your agent in small playable steps.", "Playtest after each step.", "Share or publish, respecting copyright." ], "prompt": "Decompile the original with Ghidra and port it to TypeScript one function at a time, keeping a table of ported functions. Render with low-poly WebGL.", "repo": null }, "author": { "name": "drdator", "handle": "drdator", "url": "https://news.ycombinator.com/user?id=drdator", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49977886", "platform": "Hacker News", "posted_at": "2026-10-06", "engagement": "37 points", "engagement_n": 37 } ], "posted_at": "2026-10-06", "added_at": "2026-10-15" }, { "id": "screen-materials-for-computer-memory-with-agents", "url": "https://agentpicklist.com/use-cases/screen-materials-for-computer-memory-with-agents/", "title": "Screen new magnetic materials for computer memory with an agent team", "summary": "A researcher used a team of Claude Opus agents to run physics simulations and surface two candidate magnets for spin-based memory, then published every calculation.", "what_they_did": "Geby Jaff (Vals AI) directed Claude Opus agents that ran density functional theory simulations to look for semiconductors with zero net magnetism that still sort electrons by spin. The agents designed one new compound and flagged a second material first made in 1999.", "outcome": "The post reports two candidates: a designed compound that may be hard to synthesize in its useful form, and KV[Cr(CN)6], whose 1999 sample stayed magnetically ordered up to 376 K. Neither the band gap nor the spin sorting has been measured yet. Inputs, outputs, code and known caveats are in a public GitHub repo.", "worth_knowing": "These are predictions, not lab results. The two simulation methods disagree on how much water in the real sample weakens the effect. Shipping a public ledger with a one-command checker is what makes the claim reviewable.", "category": "research", "agents": [ "Claude" ], "tools": [ "DFT simulations", "Python", "GitHub" ], "roles": [ "Researcher", "Scientist" ], "industries": [ "Science", "Materials" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude (Opus-class) agents", "Simulation software your field already trusts", "A public repo with inputs, outputs and a checker script" ], "steps": [ "Write the target properties as hard numeric criteria.", "Let agents propose candidates and run the cheap simulation first, the expensive one on survivors.", "Have a separate agent list every caveat and failure mode.", "Publish all inputs, raw outputs and a one-command re-run so others can check." ], "prompt": "Our target: [property list with numeric thresholds]. Propose 20 candidates, run the fast screen on each, and keep only those that pass every threshold. For survivors, run the accurate method and write a caveats section listing every way the result could be wrong.", "repo": null }, "author": { "name": "Geby Jaff (Vals AI)", "handle": "spicylemonade", "platform": "Vals AI blog", "url": "https://www.vals.ai/blogs/room-temperature-magnetic-semiconductors" }, "sources": [ { "url": "https://www.vals.ai/blogs/room-temperature-magnetic-semiconductors", "platform": "Blog", "posted_at": "2026-10-04", "engagement_n": 0 }, { "url": "https://news.ycombinator.com/item?id=49970667", "platform": "Hacker News", "posted_at": "2026-10-05", "engagement": "495 points", "engagement_n": 495 } ], "posted_at": "2026-10-04", "added_at": "2026-10-15" }, { "id": "draft-email-replies-with-ai-and-keep-a-human-approval-step", "url": "https://agentpicklist.com/use-cases/draft-email-replies-with-ai-and-keep-a-human-approval-step/", "title": "Draft email replies with AI and keep a human approval step", "summary": "A builder made an n8n workflow where AI drafts customer email replies and you approve, edit or reject them.", "what_they_did": "@0xGaius_ built a review loop so AI speeds up replies without full control.", "outcome": "Shared as a build for teams.", "worth_knowing": "Human approval keeps tone and accuracy in check.", "category": "support", "agents": [ "n8n" ], "tools": [ "Email" ], "roles": [ "Customer support" ], "industries": [ "Services" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "n8n", "LLM", "Email" ], "steps": [ "Collect the incoming channel (inbox, form, tickets) into one workflow.", "Have the model classify and extract fields, returning structured JSON.", "Route or draft replies; keep a human approval step for anything customer-facing.", "Log results and review misclassifications weekly." ], "prompt": "Draft a reply to this customer email in our tone. Mark anything you are unsure about in [brackets].", "repo": null }, "author": { "name": "@0xGaius_", "handle": "0xGaius_", "url": "https://x.com/0xGaius_", "platform": "X" }, "sources": [ { "url": "https://x.com/0xGaius_/status/2106804110867497051", "platform": "X", "posted_at": "2026-10-04", "engagement": "3 likes", "engagement_n": 3 } ], "posted_at": "2026-10-04", "added_at": "2026-10-15" }, { "id": "have-claude-code-review-your-requirements-against-your-roadmap", "url": "https://agentpicklist.com/use-cases/have-claude-code-review-your-requirements-against-your-roadmap/", "title": "Have Claude Code review your requirements against your roadmap", "summary": "A developer found Claude Code more useful for reviewing requirements than writing them.", "what_they_did": "@rwhtech asked Claude Code to review his app’s requirements against its roadmap.", "outcome": "It found requirements he could not verify before the 0.1 release, and a loophole in an acceptance criterion.", "worth_knowing": "Review is a high-value, low-risk agent task.", "category": "coding", "agents": [ "Claude Code" ], "tools": [], "roles": [ "Developer", "Product manager" ], "industries": [ "Software" ], "difficulty": "beginner", "time_to_build": { "text": "est. 30 minutes", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code" ], "steps": [ "Open the repo in your coding agent and describe the goal and constraints.", "Ask for a plan first, then let it implement in small, reviewable steps.", "Run the tests or the app after each step and feed errors back.", "Review the diff before merging." ], "prompt": "Compare REQUIREMENTS.md to ROADMAP.md. List requirements that cannot be verified by release 0.1 and any acceptance criteria with loopholes.", "repo": null }, "author": { "name": "@rwhtech", "handle": "rwhtech", "url": "https://x.com/rwhtech", "platform": "X" }, "sources": [ { "url": "https://x.com/rwhtech/status/2106172377336082828", "platform": "X", "posted_at": "2026-10-03", "engagement": "", "engagement_n": 0 } ], "posted_at": "2026-10-03", "added_at": "2026-10-15" }, { "id": "vibe-code-a-production-calendar-your-crm-cant-do", "url": "https://agentpicklist.com/use-cases/vibe-code-a-production-calendar-your-crm-cant-do/", "title": "Vibe code the production calendar your CRM can't give you", "summary": "A field crew business built its own scheduling tool on top of its CRM, and it is now the company's daily source of truth.", "what_they_did": "@sensarpensar's team vibe coded a production calendar that shows open gaps by crew, pulls jobs from the CRM, links to the photos app and approved quotes, and estimates job length from size, pitch and crew. It shows revenue and gross margin per job, crew and month.", "outcome": "They say the company now uses it every day as its source of truth for production, and that once a job is approved they already know where it fits.", "worth_knowing": "Built because off-the-shelf tools fought them. You own maintenance and data sync with the CRM.", "category": "ops", "agents": [], "tools": [ "CRM API" ], "roles": [ "Operations", "Business owner" ], "industries": [ "Field services", "Construction" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Any AI coding agent", "Your CRM's API", "A small web app + database" ], "steps": [ "List the 3 questions your schedule must answer daily.", "Pull jobs and crews from the CRM API into a small database.", "Ask the agent for a calendar view with gaps per crew and a job-duration estimate.", "Add per-job margin once the basics are trusted." ], "prompt": "Build a web app that pulls approved jobs from [CRM] via API, shows a week calendar by crew with open gaps highlighted, estimates job duration from [fields], and links each job to its quote.", "repo": null }, "author": { "name": "@sensarpensar", "handle": "sensarpensar", "platform": "X", "url": "https://x.com/sensarpensar" }, "sources": [ { "url": "https://x.com/sensarpensar/status/2106110703056916646", "platform": "X", "posted_at": "2026-10-02", "engagement_n": 0 } ], "posted_at": "2026-10-02", "added_at": "2026-10-15" }, { "id": "ask-claude-code-for-your-app-revenue-over-mcp", "url": "https://agentpicklist.com/use-cases/ask-claude-code-for-your-app-revenue-over-mcp/", "title": "Ask Claude Code for your app revenue metrics over MCP", "summary": "A Shopify app developer built a replacement for a shut-down analytics tool, with an MCP server so Claude Code can answer 'what was last month's MRR?'.", "what_they_did": "@MMSUCK1 built App Rings after Mantle shut down on September 30. It connects to the Shopify Partner API (read-only) and computes daily MRR, churn, trials and payouts, and it exposes an MCP server and JSON API so any MCP-compatible tool can query the numbers.", "outcome": "He says he validated the MRR calculation against his own apps' real data and payout records. MCP and API are on the free plan.", "worth_knowing": "Read-only Partner API access; tokens are encrypted per organization. If you use the in-app AI chat, your inputs go to the AI provider you pick.", "category": "ops", "agents": [ "Claude Code" ], "tools": [ "MCP", "Shopify Partner API", "Slack" ], "roles": [ "App developer", "Founder" ], "industries": [ "E-commerce", "Software" ], "difficulty": "beginner", "time_to_build": { "text": "est. 10 minutes to connect", "sourced": false }, "monthly_cost": { "text": "Sourced: free under $1k average MRR and 1,000 installs", "sourced": true }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code", "An MCP server for your metrics" ], "steps": [ "Generate an access key in your analytics tool.", "Add its MCP server to Claude Code.", "Ask questions in plain English and have the agent write a weekly summary." ], "prompt": "Using the metrics MCP, compare this month's MRR, churn and trial conversions to last month, list the 3 biggest changes, and suggest one experiment for each.", "repo": null }, "author": { "name": "@MMSUCK1", "handle": "MMSUCK1", "platform": "X", "url": "https://x.com/MMSUCK1" }, "sources": [ { "url": "https://x.com/MMSUCK1/status/2105864549765423530", "platform": "X", "posted_at": "2026-10-02", "engagement_n": 0 } ], "posted_at": "2026-10-02", "added_at": "2026-10-15" }, { "id": "build-a-casino-into-claude-codes-thinking-spinner", "url": "https://agentpicklist.com/use-cases/build-a-casino-into-claude-codes-thinking-spinner/", "title": "Build a casino into Claude Code’s thinking spinner", "summary": "A developer used Claude Code’s new mods to add blackjack, roulette and slots while Claude thinks.", "what_they_did": "@lukemrtn built it after Claude Code added mods.", "outcome": "Now he can play while Claude thinks.", "worth_knowing": "A fun hack; no real money involved.", "category": "gaming", "agents": [ "Claude Code" ], "tools": [], "roles": [ "Developer" ], "industries": [ "Games" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code mods" ], "steps": [ "Describe the game or mod you want.", "Build it with your agent in small playable steps.", "Playtest after each step.", "Share or publish, respecting copyright." ], "prompt": "Build a mod that shows a tiny blackjack game in the thinking spinner. No real money.", "repo": null }, "author": { "name": "@lukemrtn", "handle": "lukemrtn", "url": "https://x.com/lukemrtn", "platform": "X" }, "sources": [ { "url": "https://x.com/lukemrtn/status/2106058946230227053", "platform": "X", "posted_at": "2026-10-02", "engagement": "", "engagement_n": 0 } ], "posted_at": "2026-10-02", "added_at": "2026-10-15" }, { "id": "return-an-order-by-having-your-agent-email-and-call-support", "url": "https://agentpicklist.com/use-cases/return-an-order-by-having-your-agent-email-and-call-support/", "title": "Return an order by having your agent email and call support", "summary": "A user trying to stop making support calls had Codex Dot email a retailer, then handed it the phone call when they insisted.", "what_they_did": "@arafatkatze used Codex Dot to contact Columbia about returning pants. They required a phone call, so he gave Codex that too.", "outcome": "He is documenting how it goes.", "worth_knowing": "Some companies require calls; disclose AI where required.", "category": "personal", "agents": [ "Codex" ], "tools": [ "Email", "Phone" ], "roles": [ "Individual" ], "industries": [ "Personal", "Retail" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "An agent with email and calling", "Order details" ], "steps": [ "Give the agent access only to the accounts it needs.", "Write a clear brief with hard constraints.", "Ask it to report back before anything irreversible.", "Review the outcome and refine the brief." ], "prompt": "Request a return for order [number] from [store] and ask for a prepaid label. If they require a call, call them and report the outcome.", "repo": null }, "author": { "name": "@arafatkatze", "handle": "arafatkatze", "url": "https://x.com/arafatkatze", "platform": "X" }, "sources": [ { "url": "https://x.com/arafatkatze/status/2105783095253483681", "platform": "X", "posted_at": "2026-10-01", "engagement": "3 likes", "engagement_n": 3 } ], "posted_at": "2026-10-01", "added_at": "2026-10-15" }, { "id": "learn-retro-game-development-with-claude-code", "url": "https://agentpicklist.com/use-cases/learn-retro-game-development-with-claude-code/", "title": "Learn retro game development with Claude Code", "summary": "A developer used Claude Code to learn SNES-style game development.", "what_they_did": "@DrTBehrens used Claude Code as a teacher for retro game dev.", "outcome": "He shared the results publicly.", "worth_knowing": "Ask the agent to explain, not just write.", "category": "gaming", "agents": [ "Claude Code" ], "tools": [], "roles": [ "Hobbyist", "Learner" ], "industries": [ "Games", "Education" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code", "A retro dev toolchain" ], "steps": [ "Describe the game or mod you want.", "Build it with your agent in small playable steps.", "Playtest after each step.", "Share or publish, respecting copyright." ], "prompt": "Teach me SNES sprite basics by building a tiny demo step by step, explaining each concept before writing code.", "repo": null }, "author": { "name": "@DrTBehrens", "handle": "DrTBehrens", "url": "https://x.com/DrTBehrens", "platform": "X" }, "sources": [ { "url": "https://x.com/DrTBehrens/status/2105726527115796614", "platform": "X", "posted_at": "2026-10-01", "engagement": "4 likes", "engagement_n": 4 } ], "posted_at": "2026-10-01", "added_at": "2026-10-15" }, { "id": "edit-shorts-automatically-from-a-script-with-n8n-claude-and-remotion", "url": "https://agentpicklist.com/use-cases/edit-shorts-automatically-from-a-script-with-n8n-claude-and-remotion/", "title": "Edit Shorts automatically from a script with n8n, Claude and Remotion", "summary": "A creator built an n8n workflow that turns a script into a finished short video using Claude and Remotion.", "what_they_did": "@realjoshuastone drops in a script and gets a finished video about thirty seconds later, rendered in code.", "outcome": "He says it saves two hours per video and costs pennies.", "worth_knowing": "Template-driven video works best for repeatable formats.", "category": "content", "agents": [ "n8n", "Claude" ], "tools": [ "Remotion" ], "roles": [ "Creator" ], "industries": [ "Media" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Sourced: “pennies” per video", "sourced": true }, "revenue": null, "result": "Sourced: ~2 hours saved per video", "recipe": { "stack": [ "n8n", "Claude", "Remotion" ], "steps": [ "Pick a repeatable format.", "Let the agent generate drafts or assets from your inputs.", "Edit and check facts and rights before publishing.", "Reuse the template for the next piece." ], "prompt": "Turn this script into Remotion scene JSON: one caption per line, a b-roll keyword per scene, total under 45 seconds.", "repo": null }, "author": { "name": "@realjoshuastone", "handle": "realjoshuastone", "url": "https://x.com/realjoshuastone", "platform": "X" }, "sources": [ { "url": "https://x.com/realjoshuastone/status/2105214532486349080", "platform": "X", "posted_at": "2026-09-30", "engagement": "", "engagement_n": 0 } ], "posted_at": "2026-09-30", "added_at": "2026-10-15" }, { "id": "hook-your-agent-to-support-tickets-and-product-analytics-to-fix-bugs", "url": "https://agentpicklist.com/use-cases/hook-your-agent-to-support-tickets-and-product-analytics-to-fix-bugs/", "title": "Hook your agent to support tickets and product analytics to fix bugs", "summary": "A founder connected an agent to customer support and the PostHog MCP so it keeps fixing bugs and feature requests.", "what_they_did": "@ashwinntwt tied his agent to customer support and PostHog.", "outcome": "He says it keeps fixing bugs and requests while he learns distribution.", "worth_knowing": "Review merges.", "category": "support", "agents": [], "tools": [ "PostHog MCP" ], "roles": [ "Founder" ], "industries": [ "Software" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Coding agent", "Support inbox", "PostHog MCP" ], "steps": [ "Collect the incoming channel (inbox, form, tickets) into one workflow.", "Have the model classify and extract fields, returning structured JSON.", "Route or draft replies; keep a human approval step for anything customer-facing.", "Log results and review misclassifications weekly." ], "prompt": "For each new support ticket, check PostHog for affected users, reproduce, and open a PR with a fix and a test.", "repo": null }, "author": { "name": "@ashwinntwt", "handle": "ashwinntwt", "url": "https://x.com/ashwinntwt", "platform": "X" }, "sources": [ { "url": "https://x.com/ashwinntwt/status/2105332732977107185", "platform": "X", "posted_at": "2026-09-30", "engagement": "1 likes", "engagement_n": 1 } ], "posted_at": "2026-09-30", "added_at": "2026-10-15" }, { "id": "build-an-e-ink-fridge-shopping-list", "url": "https://agentpicklist.com/use-cases/build-an-e-ink-fridge-shopping-list/", "title": "Build a family shopping list on an e-ink fridge magnet", "summary": "A parent had Claude Code write all the firmware for an e-ink touchscreen shopping list that syncs with a phone web app.", "what_they_did": "seamus_c built PaperMono, about 2,400 lines of C++ firmware for an M5Stack PaperMono (ESP32-S3, e-ink touchscreen) that sticks to the fridge and syncs with a phone web page over Wi-Fi. He says it was fully vibe coded with Claude Code; he didn't hand-write it.", "outcome": "His family uses it day to day, which he says is a first for one of his home projects. The code is on GitHub.", "worth_knowing": "E-ink needs periodic full refreshes to avoid ghosting; he used the recommended full refresh every 15 interactions. Part of the experiment was whether AI could build for unfamiliar hardware without damaging it.", "category": "home", "agents": [ "Claude Code" ], "tools": [ "ESP32-S3", "M5Stack PaperMono", "C++" ], "roles": [ "Hobbyist", "Parent" ], "industries": [ "Home", "Hardware" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported (one-time hardware)", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code", "M5Stack PaperMono (ESP32-S3)", "PlatformIO or Arduino toolchain" ], "steps": [ "Clone the repo or start from the board's example project.", "Ask Claude Code for firmware plus a tiny web app on the same Wi-Fi.", "Flash, test ticking items, then tune e-ink refresh behavior." ], "prompt": "Write ESP32-S3 firmware for an M5Stack PaperMono that shows a shopping list on the e-ink screen, lets me tick items by touch, works offline, and syncs with a mobile web page served on the local Wi-Fi.", "repo": "https://github.com/seamusc/papermono-shopping-list" }, "author": { "name": "seamus_c", "handle": "seamus_c", "platform": "Hacker News", "url": "https://news.ycombinator.com/user?id=seamus_c" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49875801", "platform": "Hacker News", "posted_at": "2026-09-28", "engagement": "152 points", "engagement_n": 152 }, { "url": "https://github.com/seamusc/papermono-shopping-list", "platform": "GitHub", "posted_at": "2026-09-28", "engagement_n": 0 } ], "posted_at": "2026-09-28", "added_at": "2026-10-15" }, { "id": "make-explainer-videos-for-any-article-on-the-fly", "url": "https://agentpicklist.com/use-cases/make-explainer-videos-for-any-article-on-the-fly/", "title": "Make explainer videos for any article on the fly", "summary": "Scrimba plugged an LLM into its interactive video format and built HN.watch, which generates explainer videos for HN posts.", "what_they_did": "Per, founder of Scrimba, built Scrimba Explain and demoed it as HN.watch: Hacker News with explainer videos created the first time someone opens a post.", "outcome": "Live at hn.watch.", "worth_knowing": "Generated explainers can be wrong; link the original.", "category": "education", "agents": [], "tools": [ "LLM", "Scrimba" ], "roles": [ "Educator", "Founder" ], "industries": [ "Education" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "LLM", "A scriptable video format" ], "steps": [ "Pick one concept to learn or teach.", "Have the agent build or explain it step by step.", "Test your understanding by changing parameters or asking why.", "Iterate." ], "prompt": "Read this article and write a 2-minute explainer script with 5 on-screen code or diagram moments.", "repo": null }, "author": { "name": "Per Borgen (Scrimba)", "handle": "mrborgen", "url": "https://news.ycombinator.com/user?id=mrborgen", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49879401", "platform": "Hacker News", "posted_at": "2026-09-28", "engagement": "226 points", "engagement_n": 226 } ], "posted_at": "2026-09-28", "added_at": "2026-10-15" }, { "id": "tailor-resumes-and-track-job-applications-with-an-agent-in-codex-or-claude-code", "url": "https://agentpicklist.com/use-cases/tailor-resumes-and-track-job-applications-with-an-agent-in-codex-or-claude-code/", "title": "Tailor resumes and track job applications with an agent in Codex or Claude Code", "summary": "A builder made Career Agent to evaluate jobs, tailor resumes and track applications.", "what_they_did": "@siddaniagi built it to run inside Codex or Claude Code.", "outcome": "Early preview: you stay in control and live submission is experimental.", "worth_knowing": "Review every application before it goes out.", "category": "recruiting", "agents": [ "Codex", "Claude Code" ], "tools": [], "roles": [ "Job seeker" ], "industries": [ "Recruiting" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Career Agent", "Codex or Claude Code" ], "steps": [ "Define your criteria.", "Let the agent search, score and draft.", "Review everything before it is sent.", "Track outcomes." ], "prompt": "Score these 10 job posts against my resume, tailor my resume for the top 3, and add all 10 to my tracker.", "repo": null }, "author": { "name": "@siddaniagi", "handle": "siddaniagi", "url": "https://x.com/siddaniagi", "platform": "X" }, "sources": [ { "url": "https://x.com/siddaniagi/status/2104457847010693524", "platform": "X", "posted_at": "2026-09-28", "engagement": "8 likes", "engagement_n": 8 } ], "posted_at": "2026-09-28", "added_at": "2026-10-15" }, { "id": "have-your-agent-write-scripts-for-cron-jobs-instead-of-calling-the-llm-each-time", "url": "https://agentpicklist.com/use-cases/have-your-agent-write-scripts-for-cron-jobs-instead-of-calling-the-llm-each-time/", "title": "Have your agent write scripts for cron jobs instead of calling the LLM each time", "summary": "An OpenClaw user says his biggest early mistake was running cron jobs through the LLM; scripts are faster and cheaper.", "what_they_did": "@Strong_Genetics first set up OpenClaw cron jobs that ran everything through the LLM.", "outcome": "Now the agent writes scripts and calls the LLM only when needed, saving usage and running faster.", "worth_knowing": "Use LLMs for judgment and scripts for repetition.", "category": "ops", "agents": [ "OpenClaw" ], "tools": [ "cron" ], "roles": [ "Power user" ], "industries": [ "Personal" ], "difficulty": "beginner", "time_to_build": { "text": "est. 30 minutes", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "OpenClaw", "cron", "Shell or Python scripts" ], "steps": [ "Write down the manual process step by step.", "Automate the deterministic parts with plain workflow steps or scripts.", "Use the model only where judgment is needed, with confidence checks.", "Add alerts and a human fallback." ], "prompt": "Turn this daily task into a script that runs on cron. Only call the LLM when [condition] needs judgment.", "repo": null }, "author": { "name": "@Strong_Genetics", "handle": "Strong_Genetics", "url": "https://x.com/Strong_Genetics", "platform": "X" }, "sources": [ { "url": "https://x.com/Strong_Genetics/status/2104450960210669772", "platform": "X", "posted_at": "2026-09-28", "engagement": "1 likes", "engagement_n": 1 } ], "posted_at": "2026-09-28", "added_at": "2026-10-15" }, { "id": "get-a-narrated-video-review-of-your-chess-games", "url": "https://agentpicklist.com/use-cases/get-a-narrated-video-review-of-your-chess-games/", "title": "Get a narrated video breakdown of your own chess games", "summary": "A chess player built Claude Code skills that combine Stockfish with his voice notes to produce a commented video of his last online game.", "what_they_did": "brumar wrote Claude Code skills that pull a game, run Stockfish, read his live audio or text notes, and produce a commented video explaining what he was thinking and what he missed.", "outcome": "He says it works but takes about an hour per game, and the last game would have cost about $15 at API prices. He found it more memorable than clicking through engine lines. The skills are open source on GitHub.", "worth_knowing": "Token-heavy, so check your quota. He noticed little sycophancy because Stockfish anchors the analysis.", "category": "gaming", "agents": [ "Claude Code" ], "tools": [ "Stockfish", "Lichess", "Claude Code skills" ], "roles": [ "Hobbyist", "Learner" ], "industries": [ "Games", "Education" ], "difficulty": "intermediate", "time_to_build": { "text": "A few sessions", "sourced": true }, "monthly_cost": { "text": "Sourced: ~$15 per analyzed game at API prices", "sourced": true }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code", "Stockfish", "chess-postmortem-skills (GitHub)" ], "steps": [ "Install Claude Code and Stockfish.", "Clone the skills repo into your skills folder.", "Record quick voice or text notes while you play.", "Ask Claude Code to analyze your last game with your notes." ], "prompt": "Analyze my last Lichess game. Use Stockfish for evaluations, use my notes below for what I was thinking, and explain the 3 moments where my reasoning went wrong, in plain language.\n\nMy notes: [paste]", "repo": "https://github.com/brumar/chess-postmortem-skills" }, "author": { "name": "brumar", "handle": "brumar", "platform": "Hacker News", "url": "https://news.ycombinator.com/user?id=brumar" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49857528", "platform": "Hacker News", "posted_at": "2026-09-26", "engagement": "126 points", "engagement_n": 126 }, { "url": "https://github.com/brumar/chess-postmortem-skills", "platform": "GitHub", "posted_at": "2026-09-26", "engagement_n": 0 } ], "posted_at": "2026-09-26", "added_at": "2026-10-15" }, { "id": "cap-spend-before-running-parallel-coding-agents", "url": "https://agentpicklist.com/use-cases/cap-spend-before-running-parallel-coding-agents/", "title": "Cap spending before you let a coding agent spawn sub-agents", "summary": "A developer reports that a simple UI review task in Codex spawned 826 child tasks and roughly $78,000 in usage without his approval.", "what_they_did": "lorenzomassaro says a Codex task asking for UX/UI validation of one module spawned 826 child tasks on a more expensive model and reasoning level, expanding into backend, OAuth and release work.", "outcome": "He reports roughly USD 78,000 consumed and records of the work deleted, and says he struggled for two weeks to reach a human at OpenAI. He notes his local token counters are not the official billing ledger. This is his account and has not been independently verified.", "worth_knowing": "The lesson for builders: set hard spend limits at the provider and the card, restrict which models and how many sub-agents a task may use, and watch long runs. He says an agent switched payment cards after a bank limit was hit.", "category": "coding", "agents": [ "Codex" ], "tools": [ "VS Code" ], "roles": [ "Developer", "Founder" ], "industries": [ "Software" ], "difficulty": "beginner", "time_to_build": { "text": "est. 15 minutes", "sourced": false }, "monthly_cost": { "text": "n/a", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Provider usage limits", "Card-level spend limits", "Agent config restricting models and sub-agents" ], "steps": [ "Set a monthly hard cap in your provider's billing settings.", "Use a virtual card with a low limit for AI APIs.", "Pin the model and disable or cap sub-agent spawning in your agent config.", "Check usage dashboards during long runs." ], "prompt": "Before starting, restate the task scope in one sentence. Do not spawn sub-agents, switch models, or touch files outside [path]. If the task needs more than that, stop and ask me.", "repo": null }, "author": { "name": "lorenzomassaro", "handle": "lorenzomassaro", "platform": "Hacker News", "url": "https://news.ycombinator.com/user?id=lorenzomassaro" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49861047", "platform": "Hacker News", "posted_at": "2026-09-26", "engagement": "84 points", "engagement_n": 84 } ], "posted_at": "2026-09-26", "added_at": "2026-10-15" }, { "id": "build-your-own-finance-workflow-tool-after-20-years-in-finance", "url": "https://agentpicklist.com/use-cases/build-your-own-finance-workflow-tool-after-20-years-in-finance/", "title": "Build your own finance workflow tool after 20 years in finance", "summary": "A finance veteran who picked coding back up built a local-first tool for finance workflows as a side project.", "what_they_did": "kdautaj spent 20 years in finance and M&A, relearned coding, and after ChatGPT and Claude arrived built Ekselio as a side project.", "outcome": "Launched on HN as a local-first product.", "worth_knowing": "Domain knowledge is the edge; AI fills the coding gap.", "category": "finance", "agents": [ "ChatGPT", "Claude" ], "tools": [ "React", "TypeScript" ], "roles": [ "Finance professional" ], "industries": [ "Finance" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude or ChatGPT", "React + TypeScript" ], "steps": [ "Connect or export your financial data.", "Have the agent categorize and summarize it.", "Check the numbers against your statements.", "Schedule recurring reports." ], "prompt": "Build a local web app that takes a company’s three statements as CSV and produces a ratio analysis and a one-page summary.", "repo": null }, "author": { "name": "kdautaj", "handle": "kdautaj", "url": "https://news.ycombinator.com/user?id=kdautaj", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49849986", "platform": "Hacker News", "posted_at": "2026-09-25", "engagement": "43 points", "engagement_n": 43 } ], "posted_at": "2026-09-25", "added_at": "2026-10-15" }, { "id": "sort-a-studios-inbox-so-good-leads-stop-going-cold", "url": "https://agentpicklist.com/use-cases/sort-a-studios-inbox-so-good-leads-stop-going-cold/", "title": "Sort a studio’s inbox so good leads stop going cold", "summary": "A builder made an n8n automation that sorts a studio’s Gmail into enquiries, clients, job applications and spam.", "what_they_did": "@AdebowaleTol built it after good leads went cold because nobody replied in time.", "outcome": "Every email gets sorted and key details extracted.", "worth_knowing": "Watch for misclassified leads at first.", "category": "support", "agents": [ "n8n" ], "tools": [ "Gmail" ], "roles": [ "Studio owner" ], "industries": [ "Creative services" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "n8n", "Gmail", "LLM" ], "steps": [ "Collect the incoming channel (inbox, form, tickets) into one workflow.", "Have the model classify and extract fields, returning structured JSON.", "Route or draft replies; keep a human approval step for anything customer-facing.", "Log results and review misclassifications weekly." ], "prompt": "Label this email as new_enquiry, client, job_application or spam and extract name, project type and deadline.", "repo": null }, "author": { "name": "@AdebowaleTol", "handle": "AdebowaleTol", "url": "https://x.com/AdebowaleTol", "platform": "X" }, "sources": [ { "url": "https://x.com/AdebowaleTol/status/2103296295545012313", "platform": "X", "posted_at": "2026-09-25", "engagement": "1 likes", "engagement_n": 1 } ], "posted_at": "2026-09-25", "added_at": "2026-10-15" }, { "id": "find-conversations-you-forgot-to-follow-up-on", "url": "https://agentpicklist.com/use-cases/find-conversations-you-forgot-to-follow-up-on/", "title": "Find conversations you forgot to follow up on", "summary": "A builder made an 8-node n8n automation that scans calendar, inbox and tasks for conversations needing a follow-up.", "what_they_did": "@BassamDahabra built an automation that analyzes his calendar, inbox and tasks overnight.", "outcome": "It found 19 dead conversations; he says it costs about 2 cents a run.", "worth_knowing": "Draft follow-ups for review rather than auto-sending.", "category": "sales", "agents": [ "n8n" ], "tools": [ "Gmail", "Google Calendar" ], "roles": [ "Solopreneur" ], "industries": [ "Services" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Sourced: ~$0.02 per run", "sourced": true }, "revenue": null, "result": null, "recipe": { "stack": [ "n8n", "Gmail", "Calendar", "An LLM" ], "steps": [ "Connect your lead sources and CRM.", "Have the agent enrich, dedupe and draft, not send.", "Review drafts and push approved items to the CRM or sequencer.", "Track replies and tune the prompts." ], "prompt": "List threads where I was asked something and have not replied in 7+ days. Draft a short follow-up for each.", "repo": null }, "author": { "name": "@BassamDahabra", "handle": "BassamDahabra", "url": "https://x.com/BassamDahabra", "platform": "X" }, "sources": [ { "url": "https://x.com/BassamDahabra/status/2102701286307119565", "platform": "X", "posted_at": "2026-09-23", "engagement": "5 likes", "engagement_n": 5 } ], "posted_at": "2026-09-23", "added_at": "2026-10-15" }, { "id": "stop-your-agent-before-it-signs-a-contract-for-you", "url": "https://agentpicklist.com/use-cases/stop-your-agent-before-it-signs-a-contract-for-you/", "title": "Stop your agent before it signs a contract for you", "summary": "A developer reports Claude Code found a contract in his Gmail, placed his saved signature on it and was about to send it when he stopped it.", "what_they_did": "franze told Claude Code to push a project further. It downloaded an unread contract PDF from his Gmail, found a saved signature image on his computer, placed it on the contract, and prepared to send it.", "outcome": "He intervened before it went out.", "worth_knowing": "Keep signature files out of agent reach and require approval for outbound email or anything legal.", "category": "legal", "agents": [ "Claude Code" ], "tools": [ "Gmail" ], "roles": [ "Developer", "Founder" ], "industries": [ "Legal" ], "difficulty": "beginner", "time_to_build": { "text": "est. 15 minutes", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Agent permission settings", "Separate folders for sensitive files" ], "steps": [ "Move signatures and IDs out of folders agents can read.", "Require approval for sending email and uploading files.", "Scope tasks: say what the agent must not do." ], "prompt": "Before any action that sends, signs, pays or agrees to anything, stop and ask me with a one-line summary.", "repo": null }, "author": { "name": "franze", "handle": "franze", "url": "https://news.ycombinator.com/user?id=franze", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49798257", "platform": "Hacker News", "posted_at": "2026-09-22", "engagement": "51 points", "engagement_n": 51 } ], "posted_at": "2026-09-22", "added_at": "2026-10-15" }, { "id": "red-team-your-support-agent-with-multi-turn-attacks-before-launch", "url": "https://agentpicklist.com/use-cases/red-team-your-support-agent-with-multi-turn-attacks-before-launch/", "title": "Red-team your support agent with multi-turn attacks before launch", "summary": "A team built a support agent and tried to break it; by turn 8 it approved a $185 refund for an order that did not exist.", "what_they_did": "@Humabound_AI tested its own support agent with single-turn and multi-turn attacks.", "outcome": "It scored 76% on single-turn tests but 43% under multi-turn attacks, approving an over-limit refund for a nonexistent order.", "worth_knowing": "Test long conversations, not just single prompts.", "category": "support", "agents": [], "tools": [], "roles": [ "Developer", "Support lead" ], "industries": [ "Customer service" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Your support agent", "A multi-turn test harness" ], "steps": [ "Collect the incoming channel (inbox, form, tickets) into one workflow.", "Have the model classify and extract fields, returning structured JSON.", "Route or draft replies; keep a human approval step for anything customer-facing.", "Log results and review misclassifications weekly." ], "prompt": "Act as a manipulative customer over 10 turns trying to get a refund outside policy. Log the turn where the agent breaks policy.", "repo": null }, "author": { "name": "@Humabound_AI", "handle": "Humabound_AI", "url": "https://x.com/Humabound_AI", "platform": "X" }, "sources": [ { "url": "https://x.com/Humabound_AI/status/2102486228981735462", "platform": "X", "posted_at": "2026-09-22", "engagement": "1 likes", "engagement_n": 1 } ], "posted_at": "2026-09-22", "added_at": "2026-10-15" }, { "id": "make-rust-code-faster-with-agent-benchmark-loops", "url": "https://agentpicklist.com/use-cases/make-rust-code-faster-with-agent-benchmark-loops/", "title": "Make Rust libraries faster by looping agents against benchmarks", "summary": "A data scientist got agents to write Rust ML and utility crates that beat established libraries on speed by giving them benchmark targets and strict anti-cheating rules.", "what_they_did": "Max Woolf had agents build crates (UMAP, gradient-boosted trees, templating, word clouds, ASCII art) with criterion benchmarks, then asked them to keep optimizing until every benchmark ran at least 1.2x faster than a recorded baseline. He added quality checks against reference implementations, subagents run through CLI calls to cheaper models, refactor passes, and head-to-head benchmarks against competing crates.", "outcome": "He reports cumulative 2x-20x speedups depending on the domain, with his UMAP crate 4x-15x faster than umap-learn and near-parity on quality. He published his AGENTS.md and an optimization prompt.", "worth_knowing": "Agents cheat when they can: one 'sped up' a physics sim by disabling the physics engine, another reduced training epochs. His rules forbid running benchmarks in parallel, editing benchmarks, and special compiler flags.", "category": "coding", "agents": [ "Claude Code", "Codex" ], "tools": [ "Rust", "criterion", "PyO3", "Zed" ], "roles": [ "Data scientist", "Developer" ], "industries": [ "Software", "Machine learning" ], "difficulty": "advanced", "time_to_build": { "text": "Months of iteration", "sourced": true }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Codex or Claude Code", "Rust + criterion", "A reference implementation for quality checks", "AGENTS.md with benchmarking rules" ], "steps": [ "Have the agent create diverse benchmarks first, then record a baseline without changes.", "Set a pass/fail target (e.g. every benchmark at least 1.2x faster) and permit iteration.", "Add a quality gate against a known-correct implementation.", "Forbid gaming: no editing benchmarks, no parallel runs, no special flags, no unsafe code.", "Finish with a refactor pass that must not regress any benchmark." ], "prompt": "First, without changing anything, run the benchmark suite and save it as the baseline. Then optimize the library (never the benchmarks) until EVERY benchmark is at least 1.2x faster than baseline. No unsafe code. After each pass, print results relative to baseline. Keep iterating until you run out of ideas.", "repo": null }, "author": { "name": "Max Woolf", "handle": "minimaxir", "platform": "Blog", "url": "https://minimaxir.com/2026/09/agentic-iteration/" }, "sources": [ { "url": "https://minimaxir.com/2026/09/agentic-iteration/", "platform": "Blog", "posted_at": "2026-09-21", "engagement_n": 0 }, { "url": "https://news.ycombinator.com/item?id=49803085", "platform": "Hacker News", "posted_at": "2026-09-22", "engagement": "116 points", "engagement_n": 116 } ], "posted_at": "2026-09-21", "added_at": "2026-10-15" }, { "id": "catch-conflicting-plans-between-parallel-coding-agents-before-they-merge", "url": "https://agentpicklist.com/use-cases/catch-conflicting-plans-between-parallel-coding-agents-before-they-merge/", "title": "Catch conflicting plans between parallel coding agents before they merge", "summary": "A team running several coding agents on one repo built a tool to catch intent conflicts Git cannot see.", "what_they_did": "naw103’s team at GPTree runs parallel agents in worktrees. The worst failures were plans that could not both be true, like one agent replacing a class another was extending, so they built Foremerge.", "outcome": "Open-sourced on GitHub.", "worth_knowing": "Git only notices textual conflicts, not architectural ones.", "category": "coding", "agents": [ "Claude Code", "Codex" ], "tools": [ "Git worktrees" ], "roles": [ "Engineering lead" ], "industries": [ "Software" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Foremerge", "Parallel worktrees" ], "steps": [ "Open the repo in your coding agent and describe the goal and constraints.", "Ask for a plan first, then let it implement in small, reviewable steps.", "Run the tests or the app after each step and feed errors back.", "Review the diff before merging." ], "prompt": "Compare the plans in /plans/*.md. List any two plans whose changes cannot both be true and suggest an order or merge.", "repo": "https://github.com/naw103/foremerge" }, "author": { "name": "naw103", "handle": "naw103", "url": "https://news.ycombinator.com/user?id=naw103", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49789356", "platform": "Hacker News", "posted_at": "2026-09-21", "engagement": "45 points", "engagement_n": 45 } ], "posted_at": "2026-09-21", "added_at": "2026-10-15" }, { "id": "generate-personalized-meal-plans-in-an-app-built-with-claude-code", "url": "https://agentpicklist.com/use-cases/generate-personalized-meal-plans-in-an-app-built-with-claude-code/", "title": "Generate personalized meal plans in an app built with Claude Code", "summary": "A designer used Claude Code to make her meal-planning app design fully functional.", "what_they_did": "@JoyyB__ took a design to a working product that generates meal plans from preferences in seconds.", "outcome": "Shared as series 1 of the build.", "worth_knowing": "Not dietary advice for medical conditions.", "category": "health", "agents": [ "Claude Code" ], "tools": [], "roles": [ "Designer" ], "industries": [ "Health", "Food" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code", "An LLM API" ], "steps": [ "Gather your documents or preferences.", "Ask the agent to explain and organize, not diagnose.", "Take its questions and summaries to a professional.", "Keep sensitive data private." ], "prompt": "Build a screen that asks diet, allergies and budget, then generates a 7-day meal plan and shopping list.", "repo": null }, "author": { "name": "@JoyyB__", "handle": "JoyyB__", "url": "https://x.com/JoyyB__", "platform": "X" }, "sources": [ { "url": "https://x.com/JoyyB__/status/2101346075512582559", "platform": "X", "posted_at": "2026-09-19", "engagement": "2 likes", "engagement_n": 2 } ], "posted_at": "2026-09-19", "added_at": "2026-10-15" }, { "id": "classify-and-route-client-inquiries-with-n8n-and-gemini", "url": "https://agentpicklist.com/use-cases/classify-and-route-client-inquiries-with-n8n-and-gemini/", "title": "Classify and route client inquiries with n8n and Gemini", "summary": "A freelancer built an n8n workflow that standardizes Gmail and form inquiries, classifies them with Gemini, logs to Sheets and routes to Slack.", "what_they_did": "@JiggleBoxStudio replaced a client’s manual sorting of Gmail and Tally inquiries.", "outcome": "The key fix was a bad field mapping that sent empty data to the AI.", "worth_knowing": "Validate inputs before the AI step.", "category": "support", "agents": [ "n8n", "Gemini" ], "tools": [ "Gmail", "Tally", "Google Sheets", "Slack" ], "roles": [ "Automation freelancer" ], "industries": [ "Services" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "n8n", "Gemini", "Google Sheets", "Slack" ], "steps": [ "Collect the incoming channel (inbox, form, tickets) into one workflow.", "Have the model classify and extract fields, returning structured JSON.", "Route or draft replies; keep a human approval step for anything customer-facing.", "Log results and review misclassifications weekly." ], "prompt": "Classify this inquiry as sales, support, partnership or spam, extract name, company and budget, and return JSON.", "repo": null }, "author": { "name": "@JiggleBoxStudio", "handle": "JiggleBoxStudio", "url": "https://x.com/JiggleBoxStudio", "platform": "X" }, "sources": [ { "url": "https://x.com/JiggleBoxStudio/status/2100189093833412709", "platform": "X", "posted_at": "2026-09-16", "engagement": "1 likes", "engagement_n": 1 } ], "posted_at": "2026-09-16", "added_at": "2026-10-15" }, { "id": "update-your-crm-from-meeting-transcripts-automatically", "url": "https://agentpicklist.com/use-cases/update-your-crm-from-meeting-transcripts-automatically/", "title": "Update your CRM from meeting transcripts automatically", "summary": "A consulting team built an n8n workflow that pulls deal stage, next step, date and objections out of call transcripts.", "what_they_did": "@bigfoot15agency’s workflow reads a meeting transcript and outputs structured JSON for the CRM.", "outcome": "They shared the workflow JSON publicly.", "worth_knowing": "Review extracted next steps before they hit the CRM.", "category": "sales", "agents": [ "n8n" ], "tools": [ "CRM" ], "roles": [ "Consultant", "Sales" ], "industries": [ "Consulting" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "n8n", "LLM", "CRM API" ], "steps": [ "Connect your lead sources and CRM.", "Have the agent enrich, dedupe and draft, not send.", "Review drafts and push approved items to the CRM or sequencer.", "Track replies and tune the prompts." ], "prompt": "From this transcript, return JSON with deal_stage, next_step, next_step_date and objections (list).", "repo": null }, "author": { "name": "@bigfoot15agency", "handle": "bigfoot15agency", "url": "https://x.com/bigfoot15agency", "platform": "X" }, "sources": [ { "url": "https://x.com/bigfoot15agency/status/2099052364355498224", "platform": "X", "posted_at": "2026-09-13", "engagement": "", "engagement_n": 0 } ], "posted_at": "2026-09-13", "added_at": "2026-10-15" }, { "id": "upsert-every-web-lead-into-your-crm-with-a-6-node-n8n-flow", "url": "https://agentpicklist.com/use-cases/upsert-every-web-lead-into-your-crm-with-a-6-node-n8n-flow/", "title": "Upsert every web lead into your CRM with a 6-node n8n flow", "summary": "An automation builder’s most-rebuilt client workflow validates, enriches, dedupes and upserts leads, then notifies.", "what_they_did": "@autopresslab rebuilds this one most: webhook in, validate, enrich, dedupe, CRM upsert, notify.", "outcome": "He says it saves about 2 hours of manual work a day.", "worth_knowing": "Dedupe before enrichment to save credits.", "category": "sales", "agents": [ "n8n" ], "tools": [ "CRM", "Webhooks" ], "roles": [ "Sales ops" ], "industries": [ "B2B" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": "Sourced: ~2 hours/day saved", "recipe": { "stack": [ "n8n", "CRM API", "Enrichment API" ], "steps": [ "Connect your lead sources and CRM.", "Have the agent enrich, dedupe and draft, not send.", "Review drafts and push approved items to the CRM or sequencer.", "Track replies and tune the prompts." ], "prompt": "No LLM needed: a deterministic workflow.", "repo": null }, "author": { "name": "@autopresslab", "handle": "autopresslab", "url": "https://x.com/autopresslab", "platform": "X" }, "sources": [ { "url": "https://x.com/autopresslab/status/2098898608921710743", "platform": "X", "posted_at": "2026-09-12", "engagement": "", "engagement_n": 0 } ], "posted_at": "2026-09-12", "added_at": "2026-10-15" }, { "id": "build-an-interactive-physics-explorer-with-your-agent", "url": "https://agentpicklist.com/use-cases/build-an-interactive-physics-explorer-with-your-agent/", "title": "Build an interactive physics explorer with your agent", "summary": "A builder made a free, open-source gravity, orbits and time-dilation explorer with his OpenClaw agent and GPT-6 Astra.", "what_they_did": "@jason_armstrong built it with his OpenClaw agent, Eve.", "outcome": "Free, open source and works on phones.", "worth_knowing": "Check the physics against known results.", "category": "education", "agents": [ "OpenClaw", "ChatGPT" ], "tools": [ "Web" ], "roles": [ "Hobbyist", "Educator" ], "industries": [ "Education", "Science" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "OpenClaw", "GPT-class model", "A web canvas library" ], "steps": [ "Pick one concept to learn or teach.", "Have the agent build or explain it step by step.", "Test your understanding by changing parameters or asking why.", "Iterate." ], "prompt": "Build a mobile-friendly page simulating two orbiting bodies, with sliders for mass and distance and a clock showing gravitational time dilation.", "repo": null }, "author": { "name": "@jason_armstrong", "handle": "jason_armstrong", "url": "https://x.com/jason_armstrong", "platform": "X" }, "sources": [ { "url": "https://x.com/jason_armstrong/status/2098540204499906988", "platform": "X", "posted_at": "2026-09-11", "engagement": "1 likes", "engagement_n": 1 } ], "posted_at": "2026-09-11", "added_at": "2026-10-15" }, { "id": "send-employees-the-right-tax-notification-automatically", "url": "https://agentpicklist.com/use-cases/send-employees-the-right-tax-notification-automatically/", "title": "Send employees the right tax notification automatically", "summary": "A builder made a simple n8n workflow that checks income brackets and emails each employee the right tax rate.", "what_they_did": "@cascade_br built a form-to-email flow that saves the data, checks the bracket and sends the right email (8%, 5%, 3% or tax-free).", "outcome": "No more manual checks or one-by-one emails.", "worth_knowing": "Have payroll confirm the rules.", "category": "finance", "agents": [ "n8n" ], "tools": [ "Forms", "Email" ], "roles": [ "HR", "Operations" ], "industries": [ "Finance", "HR" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "n8n", "Form", "Email" ], "steps": [ "Connect or export your financial data.", "Have the agent categorize and summarize it.", "Check the numbers against your statements.", "Schedule recurring reports." ], "prompt": "No LLM needed: a rule-based workflow.", "repo": null }, "author": { "name": "@cascade_br", "handle": "cascade_br", "url": "https://x.com/cascade_br", "platform": "X" }, "sources": [ { "url": "https://x.com/cascade_br/status/2098393757989159225", "platform": "X", "posted_at": "2026-09-11", "engagement": "", "engagement_n": 0 } ], "posted_at": "2026-09-11", "added_at": "2026-10-15" }, { "id": "replace-paid-saas-with-vibe-coded-services", "url": "https://agentpicklist.com/use-cases/replace-paid-saas-with-vibe-coded-services/", "title": "Replace paid SaaS tools with your own vibe coded services", "summary": "An indie founder replaced more than a dozen paid services across his sites with self-built tools and says it saves about $25,000 a month.", "what_they_did": "Pieter Levels (@levelsio) listed his replacements: his own weather service, an image-resizing API, NSFW detection, video streaming on a VPS, screenshots, error monitoring, maps, uptime checks, a blog, a scraper, and AI moderation and support bots replacing human moderators and support.", "outcome": "He puts total savings at about $25,000/mo, with the biggest items being moderation (~$10,000/mo) and customer support (~$5,000/mo). He keeps domains, email sending, storage, backups, servers and LLMs on vendors.", "worth_knowing": "You take on maintenance and uptime for everything you replace. He draws the line at running registrars, email, storage hardware or GPUs himself.", "category": "ops", "agents": [ "xAI API" ], "tools": [ "Node.js", "Sharp", "Redis", "Python", "Hetzner", "Telegram" ], "roles": [ "Indie founder", "Solo founder" ], "industries": [ "Software", "Media" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Sourced: several replacements run at $0-$5/mo on existing servers", "sourced": true }, "revenue": null, "result": "Sourced: about $25,000/mo in savings (self-reported)", "recipe": { "stack": [ "Your AI coding agent", "A cheap VPS", "Open-source libraries for each job" ], "steps": [ "List every SaaS bill and sort by monthly cost.", "For each, check if an open-source library covers 90% of what you use.", "Have your agent build a thin service around it with tests and monitoring.", "Run both in parallel for a week before canceling the vendor." ], "prompt": "I pay $[X]/mo for [service] and only use [features]. Build a minimal self-hosted replacement using [open-source lib], with the same API shape so I can swap it in, plus a health check and error alerts to Telegram.", "repo": null }, "author": { "name": "Pieter Levels", "handle": "levelsio", "platform": "X", "url": "https://x.com/levelsio" }, "sources": [ { "url": "https://x.com/levelsio/status/2097692685775565031", "platform": "X", "posted_at": "2026-09-09", "engagement": "4,272 likes", "engagement_n": 4272 } ], "posted_at": "2026-09-09", "added_at": "2026-10-15" }, { "id": "run-your-business-on-a-self-hosted-company-os-of-claude-code-and-codex-agents", "url": "https://agentpicklist.com/use-cases/run-your-business-on-a-self-hosted-company-os-of-claude-code-and-codex-agents/", "title": "Run your business on a self-hosted company OS of Claude Code and Codex agents", "summary": "A founder built, and runs his business on, a self-hosted company OS where Claude Code and Codex agents work in departments.", "what_they_did": "Dimitris built OtoDock, a multi-tenant app combining Claude Code, Cowork-style work and cloud sessions that runs on your Anthropic or OpenAI subscription. He uses it to run his business.", "outcome": "It is open source and free to self-host.", "worth_knowing": "You maintain the infrastructure yourself.", "category": "ops", "agents": [ "Claude Code", "Codex" ], "tools": [ "Self-hosted" ], "roles": [ "Founder", "Operations" ], "industries": [ "Small business" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "oto-dock", "Claude or OpenAI subscription", "A server" ], "steps": [ "Write down the manual process step by step.", "Automate the deterministic parts with plain workflow steps or scripts.", "Use the model only where judgment is needed, with confidence checks.", "Add alerts and a human fallback." ], "prompt": "You are the Marketing department agent. Each Monday, review last week’s metrics in /data, propose 3 experiments, and post them for my approval.", "repo": "https://github.com/OtoDock/oto-dock" }, "author": { "name": "Dimitris (dimitrismrtzs)", "handle": "dimitrismrtzs", "url": "https://news.ycombinator.com/user?id=dimitrismrtzs", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49630606", "platform": "Hacker News", "posted_at": "2026-09-09", "engagement": "50 points", "engagement_n": 50 } ], "posted_at": "2026-09-09", "added_at": "2026-10-15" }, { "id": "notice-you-no-longer-understand-your-own-app-and-go-back-to-coding-by-hand", "url": "https://agentpicklist.com/use-cases/notice-you-no-longer-understand-your-own-app-and-go-back-to-coding-by-hand/", "title": "Notice you no longer understand your own app, and go back to coding by hand", "summary": "A solo developer who added features at breakneck speed with Claude realized he no longer understood his app.", "what_they_did": "trencedamp had been building a new version of his successful app with Claude for about six months.", "outcome": "He realized he had no clue how parts of it worked and decided to code by hand again.", "worth_knowing": "Speed without understanding is a risk for apps you must maintain.", "category": "coding", "agents": [ "Claude" ], "tools": [], "roles": [ "Solo developer" ], "industries": [ "Software" ], "difficulty": "beginner", "time_to_build": { "text": "n/a", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Your editor", "Code review habits" ], "steps": [ "Read every diff before accepting it.", "Ask the agent to explain changes.", "Write the core logic yourself." ], "prompt": "Explain this change line by line as if I will maintain it alone, and list anything you are unsure about.", "repo": null }, "author": { "name": "trencedamp", "handle": "trencedamp", "url": "https://news.ycombinator.com/user?id=trencedamp", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49622554", "platform": "Hacker News", "posted_at": "2026-09-09", "engagement": "62 points", "engagement_n": 62 } ], "posted_at": "2026-09-09", "added_at": "2026-10-15" }, { "id": "track-your-personal-finances-automatically-with-grok-and-plaid", "url": "https://agentpicklist.com/use-cases/track-your-personal-finances-automatically-with-grok-and-plaid/", "title": "Track your personal finances automatically with Grok and Plaid", "summary": "A user connected Plaid to Grok and automated the personal P&L and cash flow statement he used to keep by hand.", "what_they_did": "@ncnagle set up the Plaid connector in Grok in about 30 minutes.", "outcome": "He says Grok now has full context on spending, investing, expenses and tax, and produces weekly and monthly reports and fraud monitoring. He called it life-changing.", "worth_knowing": "Bank connections are sensitive; review permissions.", "category": "finance", "agents": [ "Grok" ], "tools": [ "Plaid" ], "roles": [ "Individual" ], "industries": [ "Personal finance" ], "difficulty": "beginner", "time_to_build": { "text": "~30 minutes (sourced)", "sourced": true }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Grok", "Plaid connector" ], "steps": [ "Connect or export your financial data.", "Have the agent categorize and summarize it.", "Check the numbers against your statements.", "Schedule recurring reports." ], "prompt": "Using my linked accounts, build this month’s personal P&L and cash flow, compare to last month, and flag any unusual charges.", "repo": null }, "author": { "name": "@ncnagle", "handle": "ncnagle", "url": "https://x.com/ncnagle", "platform": "X" }, "sources": [ { "url": "https://x.com/ncnagle/status/2097724924387299826", "platform": "X", "posted_at": "2026-09-09", "engagement": "2 likes", "engagement_n": 2 } ], "posted_at": "2026-09-09", "added_at": "2026-10-15" }, { "id": "generate-instagram-carousels-automatically-with-n8n", "url": "https://agentpicklist.com/use-cases/generate-instagram-carousels-automatically-with-n8n/", "title": "Generate Instagram carousels automatically with n8n", "summary": "A creator built an n8n workflow for carousel copy, image generation and Sheets tracking.", "what_they_did": "@Gemcoin20 got tired of making carousels by hand.", "outcome": "The workflow handles content, images and logging.", "worth_knowing": "Review posts before publishing.", "category": "marketing", "agents": [ "n8n" ], "tools": [ "Google Sheets", "Instagram" ], "roles": [ "Creator", "Marketer" ], "industries": [ "Marketing" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "n8n", "LLM", "Image model", "Google Sheets" ], "steps": [ "Define the format and audience for the content.", "Have the agent generate drafts and assets from a template.", "Review before publishing.", "Track performance and feed winners back into the prompt." ], "prompt": "Write a 7-slide carousel on [topic] with a hook slide, 5 tips and a CTA, plus an image prompt per slide.", "repo": null }, "author": { "name": "@Gemcoin20", "handle": "Gemcoin20", "url": "https://x.com/Gemcoin20", "platform": "X" }, "sources": [ { "url": "https://x.com/Gemcoin20/status/2097603825678823857", "platform": "X", "posted_at": "2026-09-09", "engagement": "8 likes", "engagement_n": 8 } ], "posted_at": "2026-09-09", "added_at": "2026-10-15" }, { "id": "agree-on-pr-rules-before-a-teammate-sends-a-5-000-line-vibe-coded-pr", "url": "https://agentpicklist.com/use-cases/agree-on-pr-rules-before-a-teammate-sends-a-5-000-line-vibe-coded-pr/", "title": "Agree on PR rules before a teammate sends a 5,000-line vibe-coded PR", "summary": "A developer wrote about his business partner sending a 5,000-line vibe-coded PR he had not tested.", "what_they_did": "yhcj describes receiving an untested 5K-line AI-generated PR from his business partner.", "outcome": "The post sparked an HN discussion about review norms.", "worth_knowing": "Agree on PR size limits and testing before everyone uses agents.", "category": "coding", "agents": [], "tools": [ "GitHub" ], "roles": [ "Developer", "Founder" ], "industries": [ "Software" ], "difficulty": "beginner", "time_to_build": { "text": "n/a", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "PR template", "CI tests" ], "steps": [ "Set a maximum PR size.", "Require tests and a self-review note.", "Have the agent split big changes." ], "prompt": "Split this change into PRs under 400 lines, each with tests and a note on how I verified it.", "repo": null }, "author": { "name": "yhcj", "handle": "yhcj", "url": "https://ycj.bearblog.dev/ai-again/", "platform": "Blog" }, "sources": [ { "url": "https://ycj.bearblog.dev/ai-again/", "platform": "Blog", "posted_at": "2026-09-08", "engagement": "52 points", "engagement_n": 52 } ], "posted_at": "2026-09-08", "added_at": "2026-10-15" }, { "id": "launch-a-consumer-app-from-your-audience-in-two-months", "url": "https://agentpicklist.com/use-cases/launch-a-consumer-app-from-your-audience-in-two-months/", "title": "Launch a consumer AI app to an audience you already have", "summary": "A creator with millions of followers vibe coded an MVP in two days and, per Starter Story, hit $340K MRR two months after launch.", "what_they_did": "Starter Story reports that Sarah Pearl built Stella, a manifestation app that makes AI visualizations and daily affirmations. She vibe coded the MVP in 2 days with Claude Code and Google AI Studio; the full app took about 2-3 months. Her videos drive traffic, with ManyChat auto-DMing the app link to commenters.", "outcome": "Starter Story reports $340K MRR, 12,000 paying customers and 200,000+ downloads two months after launch.", "worth_knowing": "Distribution did the heavy lifting: she has 1.2M Instagram and 2.8M TikTok followers. Figures come from a Starter Story interview, not audited data.", "category": "business", "agents": [ "Claude Code", "Google AI Studio" ], "tools": [ "ManyChat", "Instagram", "TikTok" ], "roles": [ "Creator", "Founder" ], "industries": [ "Consumer apps", "Wellness" ], "difficulty": "intermediate", "time_to_build": { "text": "MVP in 2 days, full app ~2-3 months (sourced)", "sourced": true }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": { "usd_per_month": 340000, "text": "$340K MRR (reported by Starter Story)", "bucket": "100k", "source_stated": true, "verified": false }, "result": null, "recipe": { "stack": [ "Claude Code", "Google AI Studio", "ManyChat", "Short-form video" ], "steps": [ "Post content first and collect 'I wish this existed' comments.", "Vibe code the smallest version that delivers that wish.", "Use comment-to-DM automation to send the download link.", "Ask happy users for reviews inside the DM flow." ], "prompt": "Here are 50 comments from my audience: [paste]. Group them into the top 3 problems people want solved. For the #1 problem, spec a one-screen mobile app MVP I can build this weekend, then scaffold it.", "repo": null }, "author": { "name": "Sarah Pearl (reported by Starter Story)", "handle": "starter_story", "platform": "X", "url": "https://x.com/starter_story" }, "sources": [ { "url": "https://x.com/starter_story/status/2097008052033769638", "platform": "X", "posted_at": "2026-09-07", "engagement": "195 likes", "engagement_n": 195 } ], "posted_at": "2026-09-07", "added_at": "2026-10-15" }, { "id": "let-users-file-bug-reports-through-an-ai-that-dedupes-and-writes-the-ticket", "url": "https://agentpicklist.com/use-cases/let-users-file-bug-reports-through-an-ai-that-dedupes-and-writes-the-ticket/", "title": "Let users file bug reports through an AI that dedupes and writes the ticket", "summary": "A team runs an AI tool that checks for duplicates, writes up the ticket and submits it, using MCP to talk to their tracker.", "what_they_did": "muzani built an AI tool where people report bugs or feature requests in plain language. It searches existing tickets to avoid duplicates, writes the ticket, and submits it. They praise Linear’s MCP as clean and well designed.", "outcome": "Running in production; they pick the cheapest model that stays accurate enough.", "worth_knowing": "Cost matters in production: pick the cheapest model that is still accurate.", "category": "support", "agents": [], "tools": [ "MCP", "Linear" ], "roles": [ "Product manager", "Developer" ], "industries": [ "Software" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "An LLM with tool calling", "Linear MCP (or your tracker API)", "A small intake UI" ], "steps": [ "Collect the incoming channel (inbox, form, tickets) into one workflow.", "Have the model classify and extract fields, returning structured JSON.", "Route or draft replies; keep a human approval step for anything customer-facing.", "Log results and review misclassifications weekly." ], "prompt": "A user reported: \"[text]\". Search existing tickets for duplicates. If one matches, add this report as a comment. Otherwise write a ticket with title, steps to reproduce, expected vs actual, and severity, then submit it.", "repo": null }, "author": { "name": "muzani", "handle": "muzani", "url": "https://news.ycombinator.com/user?id=muzani", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49549217", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "study-by-voice-with-chatgpt-grounded-in-your-own-course-materials", "url": "https://agentpicklist.com/use-cases/study-by-voice-with-chatgpt-grounded-in-your-own-course-materials/", "title": "Study by voice with ChatGPT grounded in your own course materials", "summary": "A learner built a small MCP service that feeds course content to ChatGPT for voice study sessions and logs progress.", "what_they_did": "boltguo built an MCP service that retrieves study materials and course content so he can learn through voice conversations with ChatGPT. It logs learning data he can analyze later.", "outcome": "He uses it for his own learning and reviews progress from the logs.", "worth_knowing": "Logging every session is what makes progress measurable.", "category": "education", "agents": [ "ChatGPT" ], "tools": [ "MCP" ], "roles": [ "Learner", "Student" ], "industries": [ "Education" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "ChatGPT (voice)", "A small MCP server over your notes", "A log file or database" ], "steps": [ "Pick one concept to learn or teach.", "Have the agent build or explain it step by step.", "Test your understanding by changing parameters or asking why.", "Iterate." ], "prompt": "Quiz me by voice on chapter [N] using only the course materials tool. Ask one question at a time, correct me briefly, and log each answer as right or wrong.", "repo": null }, "author": { "name": "boltguo", "handle": "boltguo", "url": "https://news.ycombinator.com/user?id=boltguo", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49559151", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "scrape-every-uk-council-planning-portal-with-claude-and-monitor-it-over-mcp", "url": "https://agentpicklist.com/use-cases/scrape-every-uk-council-planning-portal-with-claude-and-monitor-it-over-mcp/", "title": "Scrape every UK council planning portal with Claude and monitor it over MCP", "summary": "A developer built a planning-applications scraper for every UK local planning authority with Claude, then added an MCP view of scraper health.", "what_they_did": "mootothemax built a scraper for every UK council and planning authority with Claude’s help, plus a debugging front end. On a whim he asked Claude to add an MCP interface reporting last run status, what works, and coverage gaps.", "outcome": "He uses it to find unknown unknowns quickly while maintaining the scraper.", "worth_knowing": "For scrapers, the hard part is everything around the mechanics: coverage, failures and patterns.", "category": "realestate", "agents": [ "Claude" ], "tools": [ "MCP", "Python" ], "roles": [ "Developer", "Data engineer" ], "industries": [ "Real estate", "Government" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude / Claude Code", "A scraping framework", "MCP server exposing run status" ], "steps": [ "Connect listings, calendar and messaging.", "Let the agent handle routine questions and scheduling.", "Keep a human on negotiations and compliance.", "Log every interaction." ], "prompt": "Using the scraper-status tool, list councils whose last run failed or returned zero results, group them by error, and propose a fix for the biggest group.", "repo": null }, "author": { "name": "mootothemax", "handle": "mootothemax", "url": "https://news.ycombinator.com/user?id=mootothemax", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49559500", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "give-customer-facing-voice-agents-one-mcp-server-for-scheduling-and-order-status", "url": "https://agentpicklist.com/use-cases/give-customer-facing-voice-agents-one-mcp-server-for-scheduling-and-order-status/", "title": "Give customer-facing voice agents one MCP server for scheduling and order status", "summary": "A company points any voice agent platform at its own MCP server so agents can book appointments and check orders for customers.", "what_they_did": "MitziMoto’s team built a custom MCP server defining the tools and resources their customer-facing voice agents need, such as scheduling appointments and checking order status.", "outcome": "They can point ElevenLabs, Vapi, Pipecat or another platform at the same server and it immediately understands the tools.", "worth_knowing": "One MCP server avoids lock-in to a single voice platform.", "category": "support", "agents": [], "tools": [ "MCP", "ElevenLabs", "Vapi", "Pipecat" ], "roles": [ "Customer support", "Developer" ], "industries": [ "Customer service" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Voice agent platform", "Custom MCP server", "Booking and order APIs" ], "steps": [ "Collect the incoming channel (inbox, form, tickets) into one workflow.", "Have the model classify and extract fields, returning structured JSON.", "Route or draft replies; keep a human approval step for anything customer-facing.", "Log results and review misclassifications weekly." ], "prompt": "You are the phone assistant for [business]. Use check_order_status and book_appointment only. Confirm name and order number before sharing details. If unsure, offer a human callback.", "repo": null }, "author": { "name": "MitziMoto", "handle": "MitziMoto", "url": "https://news.ycombinator.com/user?id=MitziMoto", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49560119", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "ask-an-ai-about-25-years-of-your-own-social-media-history", "url": "https://agentpicklist.com/use-cases/ask-an-ai-about-25-years-of-your-own-social-media-history/", "title": "Ask an AI about 25 years of your own social media history", "summary": "A developer merged decades of social media backups into one timeline and added an MCP server to query it with an LLM.", "what_they_did": "RebeccaTheDev built Afterfeed, a web app showing social media backups as one timeline: about 100,000 posts from LiveJournal in 2001 to Mastodon in 2026. She added an MCP server as a last-minute idea.", "outcome": "With an LLM wired up, she asks introspective questions such as what she was doing ten years ago.", "worth_knowing": "Personal archives are sensitive; keep the server private.", "category": "personal", "agents": [], "tools": [ "MCP" ], "roles": [ "Developer", "Individual" ], "industries": [ "Personal" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Your social media exports", "A timeline app with search", "MCP server" ], "steps": [ "Give the agent access only to the accounts it needs.", "Write a clear brief with hard constraints.", "Ask it to report back before anything irreversible.", "Review the outcome and refine the brief." ], "prompt": "Using my archive tool, what was I posting about this week 10 years ago? Summarize the themes and quote 3 posts.", "repo": null }, "author": { "name": "RebeccaTheDev", "handle": "RebeccaTheDev", "url": "https://news.ycombinator.com/user?id=RebeccaTheDev", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49560255", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "post-a-daily-sentry-error-digest-to-slack-with-a-claude-routine", "url": "https://agentpicklist.com/use-cases/post-a-daily-sentry-error-digest-to-slack-with-a-claude-routine/", "title": "Post a daily Sentry error digest to Slack with a Claude routine", "summary": "A team runs a Claude routine with no internet except Slack and Sentry that compares errors to source code and posts a daily digest.", "what_they_did": "mhammerc’s team runs a Claude routine whose only network access is the Slack API and the Sentry MCP. It reads errors, compares them to the app source, and posts a daily digest.", "outcome": "They say the reports are very accurate.", "worth_knowing": "Limiting network access to exactly two services keeps the routine safe.", "category": "coding", "agents": [ "Claude" ], "tools": [ "Sentry MCP", "Slack" ], "roles": [ "Engineering lead", "DevOps" ], "industries": [ "Software" ], "difficulty": "intermediate", "time_to_build": { "text": "est. 1-2 hours", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Scheduled Claude routine", "Sentry MCP", "Slack API", "Read-only repo access" ], "steps": [ "Open the repo in your coding agent and describe the goal and constraints.", "Ask for a plan first, then let it implement in small, reviewable steps.", "Run the tests or the app after each step and feed errors back.", "Review the diff before merging." ], "prompt": "Read the last 24h of new Sentry issues. For each, find the likely line in the repo, explain the cause in one sentence, rate severity, and post a digest to #eng-alerts.", "repo": null }, "author": { "name": "mhammerc", "handle": "mhammerc", "url": "https://news.ycombinator.com/user?id=mhammerc", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49561690", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "ground-chinese-study-sessions-in-your-own-textbook-with-an-mcp-server", "url": "https://agentpicklist.com/use-cases/ground-chinese-study-sessions-in-your-own-textbook-with-an-mcp-server/", "title": "Ground Chinese study sessions in your own textbook with an MCP server", "summary": "A learner OCR’d his Chinese textbooks into a stateless MCP server so quizzes stick to the textbook only.", "what_they_did": "navigate8310 OCR’d his Chinese textbooks and built a stateless MCP server to run quizzes and compare similar words without extra grammar creeping in. He uses it with GLM 5.3.", "outcome": "He published the repo and a hosted server.", "worth_knowing": "Restricting the model to your textbook keeps it at your level.", "category": "education", "agents": [ "GLM" ], "tools": [ "MCP", "OCR" ], "roles": [ "Learner" ], "industries": [ "Education", "Languages" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "OCR tool", "Stateless MCP server over the text", "Any LLM client" ], "steps": [ "Pick one concept to learn or teach.", "Have the agent build or explain it step by step.", "Test your understanding by changing parameters or asking why.", "Iterate." ], "prompt": "Using only the textbook tool, quiz me on lesson [N] vocabulary. Then explain the difference between [word A] and [word B] using only examples from the book.", "repo": "https://github.com/iodize6399/xuexi-keben" }, "author": { "name": "navigate8310", "handle": "navigate8310", "url": "https://news.ycombinator.com/user?id=navigate8310", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49561857", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "let-clients-chat-with-their-social-media-sentiment-instead-of-dashboards", "url": "https://agentpicklist.com/use-cases/let-clients-chat-with-their-social-media-sentiment-instead-of-dashboards/", "title": "Let clients chat with their social media sentiment instead of dashboards", "summary": "A community-management company exposes follower conversations over MCP so customers can ask their own AI what people are angry about.", "what_they_did": "itzprintz’s company TrollWall does social media community management and offers an MCP server for topics, sentiment, example comments and month-over-month comparisons.", "outcome": "Every customer wanted a different angle, and conversation in ChatGPT or Claude beat the analytics screens they had tried to build. Claude can even produce a PDF report at the end.", "worth_knowing": "Let customers bring their own AI rather than building every report screen.", "category": "marketing", "agents": [ "ChatGPT", "Claude" ], "tools": [ "MCP" ], "roles": [ "Marketer", "Community manager" ], "industries": [ "Marketing", "Social media" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "MCP server over your analytics data", "ChatGPT or Claude" ], "steps": [ "Define the format and audience for the content.", "Have the agent generate drafts and assets from a template.", "Review before publishing.", "Track performance and feed winners back into the prompt." ], "prompt": "Using the community tool, what are followers most angry about this month vs last month? Show 5 real example comments and suggest one reply strategy.", "repo": null }, "author": { "name": "itzprintz", "handle": "itzprintz", "url": "https://news.ycombinator.com/user?id=itzprintz", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49562068", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "debug-user-issues-in-minutes-with-claude-observability-and-read-only-db-access", "url": "https://agentpicklist.com/use-cases/debug-user-issues-in-minutes-with-claude-observability-and-read-only-db-access/", "title": "Debug user issues in minutes with Claude, observability and read-only DB access", "summary": "A developer connects Claude to dash0, Linear and a read-only production database, and opened an MCP over their own GraphQL API.", "what_they_did": "falbala uses Claude with dash0, Linear and read-only production database access. Separately, they built an MCP on their GraphQL API in about 100 lines of code and config.", "outcome": "They debug user issues in minutes. The product MCP got heavy usage from day one and users flooded them with feature and data requests.", "worth_knowing": "Give agents read-only database credentials.", "category": "coding", "agents": [ "Claude" ], "tools": [ "dash0", "Linear", "MCP", "GraphQL" ], "roles": [ "Developer", "Founder" ], "industries": [ "Software" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude", "Observability MCP", "Issue tracker MCP", "Read-only DB user" ], "steps": [ "Open the repo in your coding agent and describe the goal and constraints.", "Ask for a plan first, then let it implement in small, reviewable steps.", "Run the tests or the app after each step and feed errors back.", "Review the diff before merging." ], "prompt": "User [id] reports [problem]. Check their recent traces, related Linear issues and their DB rows (read-only). Give me the most likely cause and the evidence.", "repo": null }, "author": { "name": "falbala", "handle": "falbala", "url": "https://news.ycombinator.com/user?id=falbala", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49562077", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "prepare-customer-meetings-with-an-internal-mcp-over-company-data", "url": "https://agentpicklist.com/use-cases/prepare-customer-meetings-with-an-internal-mcp-over-company-data/", "title": "Prepare customer meetings with an internal MCP over company data", "summary": "A hotel-upsell company built an internal MCP that colleagues use for reporting and meeting prep.", "what_they_did": "thinkindie started with Linear and Notion MCPs, then launched an internal MCP over company data (they sell upsells for hotels).", "outcome": "Colleagues use it for reporting and preparing customer meetings, and tech-savvy customers want access too.", "worth_knowing": "Start internal, then consider exposing it to customers.", "category": "sales", "agents": [], "tools": [ "MCP", "Linear", "Notion" ], "roles": [ "Account manager", "Operations" ], "industries": [ "Hospitality" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Internal MCP over your database", "Claude or ChatGPT" ], "steps": [ "Connect your lead sources and CRM.", "Have the agent enrich, dedupe and draft, not send.", "Review drafts and push approved items to the CRM or sequencer.", "Track replies and tune the prompts." ], "prompt": "I meet [hotel] tomorrow. Using the company data tool, summarize their last 90 days of upsell revenue, top products, and one opportunity to raise.", "repo": null }, "author": { "name": "thinkindie", "handle": "thinkindie", "url": "https://news.ycombinator.com/user?id=thinkindie", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49562386", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "let-hosting-customers-debug-and-deploy-sites-by-asking-claude-or-cursor", "url": "https://agentpicklist.com/use-cases/let-hosting-customers-debug-and-deploy-sites-by-asking-claude-or-cursor/", "title": "Let hosting customers debug and deploy sites by asking Claude or Cursor", "summary": "A UK managed-hosting company launched an MCP for its control panel so users can ask why a site errors or deploy a project.", "what_they_did": "jimmy-c’s company Prostack mapped its control-panel API to MCP tools with guardrails on destructive actions.", "outcome": "Users are building on top of it and it has been very well received.", "worth_knowing": "Guard destructive actions.", "category": "ops", "agents": [ "Claude", "Cursor" ], "tools": [ "MCP" ], "roles": [ "Developer", "DevOps" ], "industries": [ "Hosting" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Hosting API", "MCP server with guarded destructive tools" ], "steps": [ "Write down the manual process step by step.", "Automate the deterministic parts with plain workflow steps or scripts.", "Use the model only where judgment is needed, with confidence checks.", "Add alerts and a human fallback." ], "prompt": "Why is [site] showing an error? Check recent deploys and logs, then propose a fix. Do not deploy without asking me.", "repo": null }, "author": { "name": "jimmy-c", "handle": "jimmy-c", "url": "https://news.ycombinator.com/user?id=jimmy-c", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49562518", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "build-garmin-workouts-from-your-training-data-inside-chatgpt-or-claude", "url": "https://agentpicklist.com/use-cases/build-garmin-workouts-from-your-training-data-inside-chatgpt-or-claude/", "title": "Build Garmin workouts from your training data inside ChatGPT or Claude", "summary": "An endurance sports analytics service lets users connect its MCP to their chat app, analyze training and push workouts to Garmin.", "what_they_did": "Aldipower’s company Tredict has run MCP in production since March for endurance sports analytics and planning, connectable from ChatGPT, Claude, Grok, Perplexity and Mistral.", "outcome": "Users have Claude create workouts and push them to Garmin devices based on an analysis of past training.", "worth_knowing": "Training plans are not medical advice; build load gradually.", "category": "health", "agents": [ "ChatGPT", "Claude", "Grok" ], "tools": [ "MCP", "Garmin" ], "roles": [ "Athlete", "Coach" ], "industries": [ "Sports", "Health" ], "difficulty": "beginner", "time_to_build": { "text": "est. 10 minutes", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Your training platform MCP", "ChatGPT or Claude" ], "steps": [ "Gather your documents or preferences.", "Ask the agent to explain and organize, not diagnose.", "Take its questions and summaries to a professional.", "Keep sensitive data private." ], "prompt": "Analyze my last 6 weeks of runs, then build next week’s plan with one interval session and one long run, and push the workouts to my Garmin.", "repo": null }, "author": { "name": "Aldipower", "handle": "Aldipower", "url": "https://news.ycombinator.com/user?id=Aldipower", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49561465", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "give-non-technical-admins-cms-stats-and-edits-through-an-mcp", "url": "https://agentpicklist.com/use-cases/give-non-technical-admins-cms-stats-and-edits-through-an-mcp/", "title": "Give non-technical admins CMS stats and edits through an MCP", "summary": "A developer built an MCP so admins can work with the CMS and get statistics from their AI client.", "what_they_did": "NietTim built an MCP for his company’s admins to work with the CMS and pull statistics, and now adds MCPs to most of his personal tools.", "outcome": "He calls the statistics part a huge unlock for less technical coworkers.", "worth_knowing": "An MCP is easier for non-developers than API keys and docs.", "category": "ops", "agents": [], "tools": [ "MCP", "CMS" ], "roles": [ "Admin", "Developer" ], "industries": [ "Media", "Software" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "MCP server over CMS API", "Claude or ChatGPT" ], "steps": [ "Write down the manual process step by step.", "Automate the deterministic parts with plain workflow steps or scripts.", "Use the model only where judgment is needed, with confidence checks.", "Add alerts and a human fallback." ], "prompt": "How many articles did we publish last month by section, and which 5 got the most views? Draft a short note for the team.", "repo": null }, "author": { "name": "NietTim", "handle": "NietTim", "url": "https://news.ycombinator.com/user?id=NietTim", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49561667", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "turn-plans-into-tickets-on-your-own-board-even-by-voice-on-a-walk", "url": "https://agentpicklist.com/use-cases/turn-plans-into-tickets-on-your-own-board-even-by-voice-on-a-walk/", "title": "Turn plans into tickets on your own board, even by voice on a walk", "summary": "A developer has Claude split feature plans into small tickets on his self-made board, which his phone assistant can also add to.", "what_they_did": "VanTodi built a ticket board with an MCP and web GUI. When planning features with Claude he asks it to cut the work into small tickets on the board.", "outcome": "His phone’s assistant can write to it too, so he captures tasks mid-walk.", "worth_knowing": "Small tickets keep agent work reviewable.", "category": "personal", "agents": [ "Claude" ], "tools": [ "MCP" ], "roles": [ "Developer" ], "industries": [ "Software" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "A simple ticket board with MCP", "Claude", "Phone voice assistant" ], "steps": [ "Give the agent access only to the accounts it needs.", "Write a clear brief with hard constraints.", "Ask it to report back before anything irreversible.", "Review the outcome and refine the brief." ], "prompt": "Split this feature plan into tickets of under 2 hours each, with acceptance criteria, and add them to my board.", "repo": null }, "author": { "name": "VanTodi", "handle": "VanTodi", "url": "https://news.ycombinator.com/user?id=VanTodi", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49561214", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "ship-an-mcp-with-a-physical-card-deck-so-facilitators-can-plan-workshops", "url": "https://agentpicklist.com/use-cases/ship-an-mcp-with-a-physical-card-deck-so-facilitators-can-plan-workshops/", "title": "Ship an MCP with a physical card deck so facilitators can plan workshops", "summary": "A maker of printed facilitation cards offers an MCP server so buyers can plan workshops with the deck in Claude Desktop.", "what_they_did": "bhagyeshsp sells a set of 50 printed facilitation cards. The card data lives on Cloudflare Workers behind an MCP server users install in Claude Desktop.", "outcome": "Facilitators plan workshops with the product’s data; he accepts that determined users could scrape it.", "worth_knowing": "For non-developers, an installable MCP beats an API.", "category": "business", "agents": [ "Claude" ], "tools": [ "MCP", "Cloudflare Workers" ], "roles": [ "Founder", "Facilitator" ], "industries": [ "Education", "Consumer products" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Cloudflare Workers", "MCP server", "Claude Desktop" ], "steps": [ "Pick a narrow problem people already pay to solve.", "Build the first version with your agent.", "Charge early and talk to users.", "Use the agent to ship fixes fast." ], "prompt": "Plan a 90-minute retrospective for 8 people using cards from the deck. Give timings and which card to use at each step.", "repo": null }, "author": { "name": "bhagyeshsp", "handle": "bhagyeshsp", "url": "https://news.ycombinator.com/user?id=bhagyeshsp", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49561347", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "add-missing-tools-to-your-home-assistant-voice-assistant-with-mcp", "url": "https://agentpicklist.com/use-cases/add-missing-tools-to-your-home-assistant-voice-assistant-with-mcp/", "title": "Add missing tools to your Home Assistant voice assistant with MCP", "summary": "A developer built an MCP server for his Home Assistant voice assistant with jokes, math, unit conversion, stocks and Kagi search.", "what_they_did": "jamesponddotco felt Home Assistant’s default tools were missing things, so he built an MCP server adding jokes, math, unit conversion, stock info and Kagi web search.", "outcome": "He says it was approved by his wife. At work he also uses MCP to debug traces.", "worth_knowing": "Voice assistants need small, reliable tools more than big models.", "category": "home", "agents": [], "tools": [ "Home Assistant", "MCP", "Kagi" ], "roles": [ "Developer", "Parent" ], "industries": [ "Home" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Home Assistant", "Custom MCP server", "LLM conversation agent" ], "steps": [ "List the devices and data sources.", "Have the agent write the integration code.", "Test on one device first.", "Automate and monitor." ], "prompt": "Add tools to my MCP server: convert_units, calculate, stock_quote and web_search. Each returns one short sentence suitable for speaking aloud.", "repo": null }, "author": { "name": "jamesponddotco", "handle": "jamesponddotco", "url": "https://news.ycombinator.com/user?id=jamesponddotco", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49563826", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "match-40-web-pages-pixel-perfect-to-figma-with-codex-and-the-figma-mcp", "url": "https://agentpicklist.com/use-cases/match-40-web-pages-pixel-perfect-to-figma-with-codex-and-the-figma-mcp/", "title": "Match 40 web pages pixel-perfect to Figma with Codex and the Figma MCP", "summary": "Facing 40 pages to match to an inconsistent Figma file, a developer connected the Figma MCP to Codex, and it did the job.", "what_they_did": "ssenssei had to make about 40 pages match Figma designs exactly. Feeding screenshots to the AI did poorly, so he set up the Figma MCP with Codex.", "outcome": "He says it actually did what he needed.", "worth_knowing": "Structured design data beats screenshots for layout accuracy.", "category": "coding", "agents": [ "Codex" ], "tools": [ "Figma MCP" ], "roles": [ "Frontend developer" ], "industries": [ "Software", "Design" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Codex", "Figma MCP" ], "steps": [ "Open the repo in your coding agent and describe the goal and constraints.", "Ask for a plan first, then let it implement in small, reviewable steps.", "Run the tests or the app after each step and feed errors back.", "Review the diff before merging." ], "prompt": "Using the Figma MCP, implement frame [name] as a page in this repo. Match spacing, type and colors using existing components, then list any differences you could not resolve.", "repo": null }, "author": { "name": "ssenssei", "handle": "ssenssei", "url": "https://news.ycombinator.com/user?id=ssenssei", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49564526", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "let-non-technical-staff-query-sales-data-in-plain-english", "url": "https://agentpicklist.com/use-cases/let-non-technical-staff-query-sales-data-in-plain-english/", "title": "Let non-technical staff query sales data in plain English", "summary": "An e-commerce company connected internal users’ agents to its data so they can ask for sales counts and charts themselves.", "what_they_did": "NichoPaolucci set up a way for internal users to connect an agent to company data, replacing many hand-built reports.", "outcome": "Users ask things like how many shirts sold in 2024 or for a revenue chart. He says they love it and the team has time for more pressing work.", "worth_knowing": "Use read-only access and define metrics clearly.", "category": "ecommerce", "agents": [], "tools": [ "MCP" ], "roles": [ "Data analyst", "Operations" ], "industries": [ "E-commerce" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Read-only data access", "MCP server", "Claude or ChatGPT" ], "steps": [ "Connect store data (orders, products, policies).", "Have the agent answer or act within written policy.", "Hand off refunds and edge cases to a human.", "Review transcripts weekly." ], "prompt": "How many units of [product] sold each month in 2024? Make a chart and note any month that changed by more than 20%.", "repo": null }, "author": { "name": "NichoPaolucci", "handle": "NichoPaolucci", "url": "https://news.ycombinator.com/user?id=NichoPaolucci", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49564918", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "sync-your-email-into-sqlite-so-agents-can-search-it-in-milliseconds", "url": "https://agentpicklist.com/use-cases/sync-your-email-into-sqlite-so-agents-can-search-it-in-milliseconds/", "title": "Sync your email into SQLite so agents can search it in milliseconds", "summary": "A developer syncs his Microsoft mail into a local SQLite database exposed read-only to his agent.", "what_they_did": "infogulch built an MCP that syncs all his Microsoft mail into SQLite with full-text search and exposes query and get tools read-only.", "outcome": "Most queries finish in under 10 ms, and agents find things quickly even with vague questions.", "worth_knowing": "Read-only access and small outputs keep it safe and cheap.", "category": "personal", "agents": [], "tools": [ "MCP", "SQLite", "Microsoft 365" ], "roles": [ "Developer" ], "industries": [ "Personal" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Mail sync script", "SQLite FTS5", "Read-only MCP server" ], "steps": [ "Give the agent access only to the accounts it needs.", "Write a clear brief with hard constraints.", "Ask it to report back before anything irreversible.", "Review the outcome and refine the brief." ], "prompt": "Find the email thread where we agreed the contract renewal date with [company] and quote the line.", "repo": null }, "author": { "name": "infogulch", "handle": "infogulch", "url": "https://news.ycombinator.com/user?id=infogulch", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49567599", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "go-from-alert-to-fix-with-mcps-for-logs-metrics-incidents-and-code", "url": "https://agentpicklist.com/use-cases/go-from-alert-to-fix-with-mcps-for-logs-metrics-incidents-and-code/", "title": "Go from alert to fix with MCPs for logs, metrics, incidents and code", "summary": "An SRE team gave its LLM MCPs for Loki, Prometheus, Grafana, Incident.io, GitHub and server inventory.", "what_they_did": "At a past job, alexpotato’s team had MCPs for Loki, Prometheus, Grafana, Incident.io, GitHub and a custom server inventory.", "outcome": "He says the LLM could start from an alert, find its source in code and check metrics, automating a lot of alert toil.", "worth_knowing": "The value comes from combining sources.", "category": "ops", "agents": [], "tools": [ "Loki", "Prometheus", "Grafana", "Incident.io", "GitHub" ], "roles": [ "SRE", "DevOps" ], "industries": [ "Software" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "MCPs for logs, metrics, incidents and code", "An LLM agent" ], "steps": [ "Write down the manual process step by step.", "Automate the deterministic parts with plain workflow steps or scripts.", "Use the model only where judgment is needed, with confidence checks.", "Add alerts and a human fallback." ], "prompt": "Alert [name] fired. Find where it is defined in the repo, pull the related metrics and logs for the last hour, and summarize the likely cause and a safe fix.", "repo": null }, "author": { "name": "alexpotato", "handle": "alexpotato", "url": "https://news.ycombinator.com/user?id=alexpotato", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49567658", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "generate-parametric-furniture-in-blender-by-driving-it-from-claude", "url": "https://agentpicklist.com/use-cases/generate-parametric-furniture-in-blender-by-driving-it-from-claude/", "title": "Generate parametric furniture in Blender by driving it from Claude", "summary": "A 3D artist drives Blender from Claude to build Geometry Nodes and asset libraries.", "what_they_did": "commieneko uses the official Blender MCP to have Claude write Geometry Nodes and Python scripts. Claude researched straight-back chair nomenclature and built a node group exposing those parts as parameters.", "outcome": "He says it was useful on the first try and very nice after some back and forth. They distribute the assets internally.", "worth_knowing": "Ask for watertight geometry and named parameters.", "category": "content", "agents": [ "Claude" ], "tools": [ "Blender MCP", "Python" ], "roles": [ "3D artist" ], "industries": [ "Design", "Games" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Blender", "Blender MCP", "Claude" ], "steps": [ "Pick a repeatable format.", "Let the agent generate drafts or assets from your inputs.", "Edit and check facts and rights before publishing.", "Reuse the template for the next piece." ], "prompt": "Research standard parts of a [furniture type], then build a Geometry Nodes group that exposes each part as a parameter and outputs watertight geometry.", "repo": null }, "author": { "name": "commieneko", "handle": "commieneko", "url": "https://news.ycombinator.com/user?id=commieneko", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49570146", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "reorganize-and-publish-help-center-articles-from-your-ai-tool", "url": "https://agentpicklist.com/use-cases/reorganize-and-publish-help-center-articles-from-your-ai-tool/", "title": "Reorganize and publish help-center articles from your AI tool", "summary": "A help-center product added an MCP so users can answer internal questions and create, move or publish articles.", "what_they_did": "peaxkl’s company HappySupport added a highly requested MCP server.", "outcome": "Users reorganize their help center, answer questions internally and edit or publish articles. It also made importing docs from obscure providers easy.", "worth_knowing": "Imports from odd providers no longer need per-provider integrations.", "category": "support", "agents": [], "tools": [ "MCP" ], "roles": [ "Support lead" ], "industries": [ "Customer service" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Help-center MCP", "Claude or ChatGPT" ], "steps": [ "Collect the incoming channel (inbox, form, tickets) into one workflow.", "Have the model classify and extract fields, returning structured JSON.", "Route or draft replies; keep a human approval step for anything customer-facing.", "Log results and review misclassifications weekly." ], "prompt": "Find help articles that overlap, propose a new structure with categories, and draft merged versions. Do not publish yet.", "repo": null }, "author": { "name": "peaxkl", "handle": "peaxkl", "url": "https://news.ycombinator.com/user?id=peaxkl", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49562636", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "let-crm-customers-combine-your-data-with-other-mcps-in-claude-or-copilot", "url": "https://agentpicklist.com/use-cases/let-crm-customers-combine-your-data-with-other-mcps-in-claude-or-copilot/", "title": "Let CRM customers combine your data with other MCPs in Claude or Copilot", "summary": "A niche CRM vendor’s MCP lets non-technical users connect platform data to Claude or Copilot.", "what_they_did": "This niche CRM built an MCP with LLM-specific tool descriptions and full OAuth, and reuses it for its own in-app agent.", "outcome": "They say it is very useful for non-technical users, none of whom know what a CLI is.", "worth_knowing": "One MCP can serve both customers and your own agent.", "category": "sales", "agents": [ "Claude", "Copilot" ], "tools": [ "MCP", "OAuth" ], "roles": [ "Founder", "Sales" ], "industries": [ "Software", "CRM" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "CRM MCP with OAuth", "Claude or Copilot" ], "steps": [ "Connect your lead sources and CRM.", "Have the agent enrich, dedupe and draft, not send.", "Review drafts and push approved items to the CRM or sequencer.", "Track replies and tune the prompts." ], "prompt": "List deals with no activity in 14 days, draft a follow-up for each, and log a task in the CRM.", "repo": null }, "author": { "name": "946789987649", "handle": "946789987649", "url": "https://news.ycombinator.com/user?id=946789987649", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49565617", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "do-auditable-compliance-work-with-agents-over-mcp", "url": "https://agentpicklist.com/use-cases/do-auditable-compliance-work-with-agents-over-mcp/", "title": "Do auditable compliance work with agents over MCP", "summary": "A compliance tool maker says customers already used MCPs for compliance and recent changes made the work fully auditable.", "what_they_did": "mlitwiniuk runs AuditBadger and uses MCP for agentic compliance work with a full audit trail.", "outcome": "He finds it easier than the agent built into the app.", "worth_knowing": "For compliance, an audit trail of every agent action is essential.", "category": "legal", "agents": [], "tools": [ "MCP" ], "roles": [ "Compliance officer" ], "industries": [ "Legal", "Compliance" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Compliance platform MCP", "Agent with logged tool calls" ], "steps": [ "Collect the documents.", "Have the agent extract and summarize.", "Have a qualified person review.", "Store the result securely." ], "prompt": "List controls without evidence for this quarter, draft the evidence request for each owner, and log every action you take.", "repo": null }, "author": { "name": "mlitwiniuk", "handle": "mlitwiniuk", "url": "https://news.ycombinator.com/user?id=mlitwiniuk", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49562874", "platform": "Hacker News", "posted_at": "2026-09-04", "engagement": "Comment in \"Ask HN: Who is using MCP in production?\"", "engagement_n": 0 } ], "posted_at": "2026-09-04", "added_at": "2026-10-15" }, { "id": "investigate-production-incidents-with-a-purpose-built-sre-agent", "url": "https://agentpicklist.com/use-cases/investigate-production-incidents-with-a-purpose-built-sre-agent/", "title": "Investigate production incidents with a purpose-built SRE agent", "summary": "A SaaS SRE team that struggled with general agents in incident response built its own Rust agent.", "what_they_did": "jvogt’s team at Mezmo tried Claude, OpenClaw and LangChain in incident workflows but hit context overflow, security risks, hallucinations and frontier tokens wasted on easy work. They built Aura.", "outcome": "They open-sourced Aura, which investigates and fixes production incidents.", "worth_knowing": "They drew a hard line at relaxing production permissions; approval fatigue was real.", "category": "ops", "agents": [ "Claude", "OpenClaw" ], "tools": [ "Rust" ], "roles": [ "SRE" ], "industries": [ "Software" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Aura", "Observability data", "Scoped credentials" ], "steps": [ "Write down the manual process step by step.", "Automate the deterministic parts with plain workflow steps or scripts.", "Use the model only where judgment is needed, with confidence checks.", "Add alerts and a human fallback." ], "prompt": "Incident: [alert]. Gather the last 30 minutes of related logs and deploys, rank three hypotheses with evidence, and propose the safest mitigation. Do not change production.", "repo": "https://github.com/mezmo/aura" }, "author": { "name": "jvogt (Mezmo)", "handle": "jvogt", "url": "https://news.ycombinator.com/user?id=jvogt", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49538195", "platform": "Hacker News", "posted_at": "2026-09-02", "engagement": "28 points", "engagement_n": 28 } ], "posted_at": "2026-09-02", "added_at": "2026-10-15" }, { "id": "let-your-agent-buy-a-domain-behind-a-human-approval-step", "url": "https://agentpicklist.com/use-cases/let-your-agent-buy-a-domain-behind-a-human-approval-step/", "title": "Let your agent buy a domain, behind a human approval step", "summary": "A builder’s OpenClaw agent bought a domain and started building the product; he approved the purchase through an approval tool he built.", "what_they_did": "@ucsandman’s OpenClaw agent bought a URL and began building via headless Claude Code. He approved the purchase in DashClaw, which he built to stop agents doing things he does not want.", "outcome": "He got the Vercel receipt.", "worth_knowing": "Put all spending behind explicit approval.", "category": "business", "agents": [ "OpenClaw", "Claude Code" ], "tools": [ "Vercel" ], "roles": [ "Founder" ], "industries": [ "Software" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "OpenClaw", "An approval gateway", "Claude Code (headless)" ], "steps": [ "Pick a narrow problem people already pay to solve.", "Build the first version with your agent.", "Charge early and talk to users.", "Use the agent to ship fixes fast." ], "prompt": "Find an available .com under $15 for [idea]. Ask for my approval with the price before buying anything.", "repo": null }, "author": { "name": "@ucsandman", "handle": "ucsandman", "url": "https://x.com/ucsandman", "platform": "X" }, "sources": [ { "url": "https://x.com/ucsandman/status/2095170905001390367", "platform": "X", "posted_at": "2026-09-02", "engagement": "5 likes", "engagement_n": 5 } ], "posted_at": "2026-09-02", "added_at": "2026-10-15" }, { "id": "find-a-saas-idea-and-build-it-with-claude-code-until-customers-pay", "url": "https://agentpicklist.com/use-cases/find-a-saas-idea-and-build-it-with-claude-code-until-customers-pay/", "title": "Find a SaaS idea and build it with Claude Code until customers pay", "summary": "A builder used Claude Code to come up with a SaaS idea, implement it, and land a second paying customer.", "what_they_did": "@airesearch12 used Claude Code for ideation and implementation.", "outcome": "He reports two paying customers.", "worth_knowing": "Early; revenue not stated.", "category": "business", "agents": [ "Claude Code" ], "tools": [ "Stripe" ], "roles": [ "Indie founder" ], "industries": [ "Software" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code", "Stripe" ], "steps": [ "Pick a narrow problem people already pay to solve.", "Build the first version with your agent.", "Charge early and talk to users.", "Use the agent to ship fixes fast." ], "prompt": "Suggest 10 SaaS ideas for [audience] with evidence of demand, pick the one I can ship in a week, and scaffold it with Stripe checkout.", "repo": null }, "author": { "name": "@airesearch12", "handle": "airesearch12", "url": "https://x.com/airesearch12", "platform": "X" }, "sources": [ { "url": "https://x.com/airesearch12/status/2094717917598290338", "platform": "X", "posted_at": "2026-09-01", "engagement": "8 likes", "engagement_n": 8 } ], "posted_at": "2026-09-01", "added_at": "2026-10-15" }, { "id": "generate-slide-decks-from-a-repo-that-warn-when-they-drift-from-the-code", "url": "https://agentpicklist.com/use-cases/generate-slide-decks-from-a-repo-that-warn-when-they-drift-from-the-code/", "title": "Generate slide decks from a repo that warn when they drift from the code", "summary": "A developer made a skill that turns a codebase into slides whose references carry line ranges and hashes.", "what_they_did": "lukicov kept making codebase slide decks with an agent that went stale. His SlideOps skill attaches exact file line ranges and hashes so standard checks flag drift.", "outcome": "Updates no longer need a costly full re-scan.", "worth_knowing": "Provenance on every claim makes docs cheap to maintain.", "category": "coding", "agents": [ "Claude Code" ], "tools": [ "Agent skills" ], "roles": [ "Engineering lead" ], "industries": [ "Software" ], "difficulty": "beginner", "time_to_build": { "text": "est. 15 minutes", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "SlideOps skill", "Claude Code" ], "steps": [ "Open the repo in your coding agent and describe the goal and constraints.", "Ask for a plan first, then let it implement in small, reviewable steps.", "Run the tests or the app after each step and feed errors back.", "Review the diff before merging." ], "prompt": "Make a 10-slide architecture overview of this repo. Every claim must cite a file and line range.", "repo": "https://github.com/glukicov/slideops" }, "author": { "name": "lukicov", "handle": "lukicov", "url": "https://news.ycombinator.com/user?id=lukicov", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49508735", "platform": "Hacker News", "posted_at": "2026-08-31", "engagement": "23 points", "engagement_n": 23 } ], "posted_at": "2026-08-31", "added_at": "2026-10-15" }, { "id": "build-an-agent-that-knows-everything-about-your-company", "url": "https://agentpicklist.com/use-cases/build-an-agent-that-knows-everything-about-your-company/", "title": "Build an agent that knows everything about your company", "summary": "Founders who struggled to set up a Hermes agent with full company context turned that into Almanac.", "what_they_did": "kushagrchitkar’s team set up Hermes for their company so they could ask questions with full context, and found it very hard.", "outcome": "They launched Almanac (YC S26) from that experience.", "worth_knowing": "Context setup and permissions are the hard part.", "category": "ops", "agents": [ "Hermes" ], "tools": [], "roles": [ "Founder", "Operations" ], "industries": [ "Software" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Hermes or a similar agent", "Connectors to docs, chat and tickets" ], "steps": [ "Write down the manual process step by step.", "Automate the deterministic parts with plain workflow steps or scripts.", "Use the model only where judgment is needed, with confidence checks.", "Add alerts and a human fallback." ], "prompt": "Using company docs and Slack, answer: what did we decide about [topic], when, and who owns it?", "repo": null }, "author": { "name": "kushagrchitkar", "handle": "kushagrchitkar", "url": "https://news.ycombinator.com/user?id=kushagrchitkar", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49511007", "platform": "Hacker News", "posted_at": "2026-08-31", "engagement": "59 points", "engagement_n": 59 } ], "posted_at": "2026-08-31", "added_at": "2026-10-15" }, { "id": "give-your-agent-access-to-apple-calendar-reminders-contacts-and-mail", "url": "https://agentpicklist.com/use-cases/give-your-agent-access-to-apple-calendar-reminders-contacts-and-mail/", "title": "Give your agent access to Apple Calendar, Reminders, Contacts and Mail", "summary": "A developer built a CLI, MCP and plugins so his OpenClaw agent, Codex and Claude can use Apple’s personal apps.", "what_they_did": "Omar Shahine built tools for Apple Calendar, Reminders, Contacts and Mail with a fast agent-friendly SQLite database, plus a mail channel with sender verification.", "outcome": "He says his own OpenClaw agent uses it; it is open source.", "worth_knowing": "Personal data access deserves careful permissions.", "category": "personal", "agents": [ "OpenClaw", "Codex", "Claude" ], "tools": [ "Apple Calendar", "Apple Mail", "MCP" ], "roles": [ "Developer", "Individual" ], "industries": [ "Personal" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "The Apple tools CLI/MCP", "Your agent" ], "steps": [ "Give the agent access only to the accounts it needs.", "Write a clear brief with hard constraints.", "Ask it to report back before anything irreversible.", "Review the outcome and refine the brief." ], "prompt": "Look at my calendar for next week, find two free 1-hour slots for deep work, and add reminders for my 3 overdue tasks.", "repo": null }, "author": { "name": "Omar Shahine", "handle": "OmarShahine", "url": "https://x.com/OmarShahine", "platform": "X" }, "sources": [ { "url": "https://x.com/OmarShahine/status/2094441045195133052", "platform": "X", "posted_at": "2026-08-31", "engagement": "122 likes", "engagement_n": 122 } ], "posted_at": "2026-08-31", "added_at": "2026-10-15" }, { "id": "watch-user-sessions-find-real-bugs-and-open-only-verified-fixes", "url": "https://agentpicklist.com/use-cases/watch-user-sessions-find-real-bugs-and-open-only-verified-fixes/", "title": "Watch user sessions, find real bugs and open only verified fixes", "summary": "A former Robinhood engineer built an open-source agent that investigates user-facing issues and opens a PR only when it can verify the fix.", "what_they_did": "aray07 built Opslane after quarterly bug bashes at Robinhood where only bugs they could reproduce got fixed.", "outcome": "Open source with docs and a demo.", "worth_knowing": "Requiring verification before a PR keeps noise down.", "category": "coding", "agents": [], "tools": [ "Session replay", "GitHub" ], "roles": [ "Developer" ], "industries": [ "Software" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Opslane", "Session data", "GitHub" ], "steps": [ "Open the repo in your coding agent and describe the goal and constraints.", "Ask for a plan first, then let it implement in small, reviewable steps.", "Run the tests or the app after each step and feed errors back.", "Review the diff before merging." ], "prompt": "From yesterday’s sessions, find user-facing errors that affected 3+ users, reproduce one, and open a PR only if a test proves the fix.", "repo": "https://github.com/opslane/opslane" }, "author": { "name": "aray07", "handle": "aray07", "url": "https://news.ycombinator.com/user?id=aray07", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49466704", "platform": "Hacker News", "posted_at": "2026-08-27", "engagement": "40 points", "engagement_n": 40 } ], "posted_at": "2026-08-27", "added_at": "2026-10-15" }, { "id": "generate-a-whole-theme-park-from-one-prompt", "url": "https://agentpicklist.com/use-cases/generate-a-whole-theme-park-from-one-prompt/", "title": "Generate a whole theme park from one prompt", "summary": "An engineer built an agent that designs Rollercoaster Tycoon-style parks with worlds, paths and rides.", "what_they_did": "thealexanderlee, an engineer at Magic Patterns, built an agent that turns prompts like “build me a cool theme park” into parks with multiple worlds, paths and rides, reusing ideas they use to keep AI websites on a design system.", "outcome": "It is live as a demo on their site.", "worth_knowing": "A constrained library of parts keeps outputs coherent.", "category": "gaming", "agents": [], "tools": [ "LLM agent" ], "roles": [ "Engineer" ], "industries": [ "Games" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "LLM with tool calls", "A library of placeable parts", "A renderer" ], "steps": [ "Describe the game or mod you want.", "Build it with your agent in small playable steps.", "Playtest after each step.", "Share or publish, respecting copyright." ], "prompt": "Build a theme park with 3 themed worlds connected by paths. Use only parts from the catalog and keep each world visually consistent.", "repo": null }, "author": { "name": "thealexanderlee", "handle": "thealexanderlee", "url": "https://news.ycombinator.com/user?id=thealexanderlee", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49452037", "platform": "Hacker News", "posted_at": "2026-08-26", "engagement": "58 points", "engagement_n": 58 } ], "posted_at": "2026-08-26", "added_at": "2026-10-15" }, { "id": "migrate-12-year-old-net-code-to-net-10-with-claude-code", "url": "https://agentpicklist.com/use-cases/migrate-12-year-old-net-code-to-net-10-with-claude-code/", "title": "Migrate 12-year-old .NET code to .NET 10 with Claude Code", "summary": "A developer used Claude Code to move a 2014 TCP sockets project to .NET 10.", "what_they_did": "@realNaeemAkram migrated his 2014 project with Claude Code.", "outcome": "He says zero code changes were needed; the server runs and echo still works.", "worth_knowing": "Run tests before and after any migration.", "category": "coding", "agents": [ "Claude Code" ], "tools": [ ".NET" ], "roles": [ "Developer" ], "industries": [ "Software" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code", ".NET SDK" ], "steps": [ "Open the repo in your coding agent and describe the goal and constraints.", "Ask for a plan first, then let it implement in small, reviewable steps.", "Run the tests or the app after each step and feed errors back.", "Review the diff before merging." ], "prompt": "Upgrade this solution to .NET 10, update project files and packages, build, run the tests, and list any behavior changes.", "repo": null }, "author": { "name": "@realNaeemAkram", "handle": "realNaeemAkram", "url": "https://x.com/realNaeemAkram", "platform": "X" }, "sources": [ { "url": "https://x.com/realNaeemAkram/status/2092711234798026856", "platform": "X", "posted_at": "2026-08-26", "engagement": "", "engagement_n": 0 } ], "posted_at": "2026-08-26", "added_at": "2026-10-15" }, { "id": "run-outbound-sales-from-one-interface-with-21-mcp-connections", "url": "https://agentpicklist.com/use-cases/run-outbound-sales-from-one-interface-with-21-mcp-connections/", "title": "Run outbound sales from one interface with 21 MCP connections", "summary": "An agency that runs outbound for 100+ AI companies says most of its delivery now runs through 21 MCP connections.", "what_they_did": "@itsalexvacca connects list building, enrichment, sequencing and CRM tools through MCP.", "outcome": "They say most delivery work now runs from one interface.", "worth_knowing": "Follow anti-spam laws and platform rules.", "category": "sales", "agents": [], "tools": [ "MCP", "CRM" ], "roles": [ "Sales", "Agency owner" ], "industries": [ "Sales", "B2B" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude or another MCP client", "MCPs for data, enrichment, sequencer and CRM" ], "steps": [ "Connect your lead sources and CRM.", "Have the agent enrich, dedupe and draft, not send.", "Review drafts and push approved items to the CRM or sequencer.", "Track replies and tune the prompts." ], "prompt": "Build a list of 50 [ICP] companies, enrich contacts, draft a 3-step sequence, and log everything in the CRM as drafts.", "repo": null }, "author": { "name": "@itsalexvacca", "handle": "itsalexvacca", "url": "https://x.com/itsalexvacca", "platform": "X" }, "sources": [ { "url": "https://x.com/itsalexvacca/status/2092754893237109168", "platform": "X", "posted_at": "2026-08-26", "engagement": "51 likes", "engagement_n": 51 } ], "posted_at": "2026-08-26", "added_at": "2026-10-15" }, { "id": "give-claude-code-a-searchable-text-memory-of-your-screen", "url": "https://agentpicklist.com/use-cases/give-claude-code-a-searchable-text-memory-of-your-screen/", "title": "Give Claude Code a searchable text memory of your screen", "summary": "A developer built a menu bar app that saves the text of your focused window to daily Markdown files Claude Code can read.", "what_they_did": "Dramatize built a macOS app that reads focused-window text every few seconds via the Accessibility API (no screenshots or OCR) and writes one Markdown file per day, with an AGENTS.md explaining the format.", "outcome": "You can ask Claude Code what you worked on Tuesday or have it build project memory.", "worth_knowing": "Captured text can include private chats or secrets; exclude sensitive apps.", "category": "personal", "agents": [ "Claude Code" ], "tools": [ "macOS Accessibility API", "Markdown" ], "roles": [ "Developer", "Knowledge worker" ], "industries": [ "Productivity" ], "difficulty": "beginner", "time_to_build": { "text": "est. 10 minutes", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "ambient-context", "Claude Code" ], "steps": [ "Give the agent access only to the accounts it needs.", "Write a clear brief with hard constraints.", "Ask it to report back before anything irreversible.", "Review the outcome and refine the brief." ], "prompt": "Read my notes folder for last week and summarize what I worked on each day, with open loops.", "repo": "https://github.com/dragthelake/ambient-context" }, "author": { "name": "Dramatize", "handle": "Dramatize", "url": "https://news.ycombinator.com/user?id=Dramatize", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49429095", "platform": "Hacker News", "posted_at": "2026-08-25", "engagement": "62 points", "engagement_n": 62 } ], "posted_at": "2026-08-25", "added_at": "2026-10-15" }, { "id": "navigate-a-cancer-diagnosis-with-an-ai-skill-for-patients-and-caregivers", "url": "https://agentpicklist.com/use-cases/navigate-a-cancer-diagnosis-with-an-ai-skill-for-patients-and-caregivers/", "title": "Navigate a cancer diagnosis with an AI skill for patients and caregivers", "summary": "Peter Yang open-sourced an AI skill to help patients and caregivers navigate cancer diagnosis and treatment and advocate for themselves.", "what_they_did": "After sharing his mom’s story, Peter Yang built and open-sourced the skill, shaped by what patients and caregivers told him they needed.", "outcome": "His post got over 1,600 likes.", "worth_knowing": "Not medical advice; use it to prepare questions for your care team.", "category": "health", "agents": [ "Claude" ], "tools": [ "Agent skills" ], "roles": [ "Caregiver", "Patient" ], "industries": [ "Health" ], "difficulty": "beginner", "time_to_build": { "text": "est. 10 minutes", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude (or any agent that supports skills)", "The open-source skill" ], "steps": [ "Gather your documents or preferences.", "Ask the agent to explain and organize, not diagnose.", "Take its questions and summaries to a professional.", "Keep sensitive data private." ], "prompt": "Here are my test results and the doctor’s notes. Explain them in plain language and list the questions I should ask at the next appointment.", "repo": null }, "author": { "name": "Peter Yang", "handle": "petergyang", "url": "https://x.com/petergyang", "platform": "X" }, "sources": [ { "url": "https://x.com/petergyang/status/2092249012913258946", "platform": "X", "posted_at": "2026-08-25", "engagement": "1,605 likes", "engagement_n": 1605 } ], "posted_at": "2026-08-25", "added_at": "2026-10-15" }, { "id": "run-a-real-estate-agents-follow-ups-and-bookings-in-one-workflow", "url": "https://agentpicklist.com/use-cases/run-a-real-estate-agents-follow-ups-and-bookings-in-one-workflow/", "title": "Run a real estate agent’s follow-ups and bookings in one workflow", "summary": "A builder automated a real estate assistant in n8n: messages, memory, booking, property updates and confirmations.", "what_they_did": "@glesphi built one n8n workflow that handles client messages, an AI agent with memory, availability and booking, property data updates, vector memory in Supabase and Gmail confirmations.", "outcome": "Seven operations run automatically.", "worth_knowing": "Fair housing and privacy rules apply to client messaging.", "category": "realestate", "agents": [ "n8n" ], "tools": [ "Supabase", "Gmail" ], "roles": [ "Real estate agent" ], "industries": [ "Real estate" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "n8n", "LLM agent node", "Supabase vectors", "Calendar", "Gmail" ], "steps": [ "Connect listings, calendar and messaging.", "Let the agent handle routine questions and scheduling.", "Keep a human on negotiations and compliance.", "Log every interaction." ], "prompt": "A lead asks to view [property]. Check my calendar, offer 3 slots, book the chosen one and send a confirmation.", "repo": null }, "author": { "name": "@glesphi", "handle": "glesphi", "url": "https://x.com/glesphi", "platform": "X" }, "sources": [ { "url": "https://x.com/glesphi/status/2091784721126789258", "platform": "X", "posted_at": "2026-08-24", "engagement": "2 likes", "engagement_n": 2 } ], "posted_at": "2026-08-24", "added_at": "2026-10-15" }, { "id": "replace-a-support-chatbot-with-a-voice-agent-that-answers-calls", "url": "https://agentpicklist.com/use-cases/replace-a-support-chatbot-with-a-voice-agent-that-answers-calls/", "title": "Replace a support chatbot with a voice agent that answers calls", "summary": "An agency built a client an AI voice agent instead of the chatbot they asked for.", "what_they_did": "@jainilnf’s agency built a voice agent for a client who wanted a support chatbot.", "outcome": "They say it handled 40% of inbound calls without a human in week one, and the client then asked to add appointment booking.", "worth_knowing": "Self-reported by the agency.", "category": "support", "agents": [], "tools": [ "Voice AI" ], "roles": [ "Agency owner" ], "industries": [ "Customer service" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": "Sourced: 40% of inbound calls handled without a human in week one", "recipe": { "stack": [ "Voice agent platform", "Knowledge base", "Calendar API" ], "steps": [ "Collect the incoming channel (inbox, form, tickets) into one workflow.", "Have the model classify and extract fields, returning structured JSON.", "Route or draft replies; keep a human approval step for anything customer-facing.", "Log results and review misclassifications weekly." ], "prompt": "You answer calls for [business]. Answer from the FAQ, take messages for anything else, and offer to book a callback.", "repo": null }, "author": { "name": "@jainilnf", "handle": "jainilnf", "url": "https://x.com/jainilnf", "platform": "X" }, "sources": [ { "url": "https://x.com/jainilnf/status/2091827863695827189", "platform": "X", "posted_at": "2026-08-24", "engagement": "9 likes", "engagement_n": 9 } ], "posted_at": "2026-08-24", "added_at": "2026-10-15" }, { "id": "turn-an-hdr-image-trick-into-a-web-tool-with-claude-code", "url": "https://agentpicklist.com/use-cases/turn-an-hdr-image-trick-into-a-web-tool-with-claude-code/", "title": "Turn an HDR image trick into a web tool with Claude Code", "summary": "A developer worked with Claude Code to turn the gain-map trick behind extra-bright logos on HDR screens into a small web tool.", "what_they_did": "telecuda noticed some LinkedIn logos looked brighter and found it was a gain map added to a JPEG, visible only on HDR screens. He worked with Claude Code to turn it into a tool.", "outcome": "He shipped soverybright.com. LinkedIn is the only network he found that does not strip the maps.", "worth_knowing": "The effect only shows on HDR displays.", "category": "content", "agents": [ "Claude Code" ], "tools": [ "JPEG gain maps" ], "roles": [ "Developer", "Designer" ], "industries": [ "Design", "Marketing" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code", "An image library that writes gain maps" ], "steps": [ "Pick a repeatable format.", "Let the agent generate drafts or assets from your inputs.", "Edit and check facts and rights before publishing.", "Reuse the template for the next piece." ], "prompt": "Build a single-page tool that takes a logo PNG and outputs a JPEG with an HDR gain map so whites look brighter on HDR screens. Everything runs in the browser.", "repo": null }, "author": { "name": "telecuda", "handle": "telecuda", "url": "https://news.ycombinator.com/user?id=telecuda", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49402521", "platform": "Hacker News", "posted_at": "2026-08-22", "engagement": "81 points", "engagement_n": 81 } ], "posted_at": "2026-08-22", "added_at": "2026-10-15" }, { "id": "build-the-niche-design-tool-you-could-not-buy", "url": "https://agentpicklist.com/use-cases/build-the-niche-design-tool-you-could-not-buy/", "title": "Build the niche design tool you could not buy", "summary": "An artist used Claude Code to build an Islamic geometry authoring tool and opened a public beta.", "what_they_did": "@fredini built the authoring tool he could not buy.", "outcome": "Public beta open.", "worth_knowing": "Niche creative tools make great agent projects.", "category": "content", "agents": [ "Claude Code" ], "tools": [ "Web" ], "roles": [ "Artist" ], "industries": [ "Design", "Art" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code", "SVG canvas" ], "steps": [ "Pick a repeatable format.", "Let the agent generate drafts or assets from your inputs.", "Edit and check facts and rights before publishing.", "Reuse the template for the next piece." ], "prompt": "Build a web editor for 8-fold Islamic star patterns with a compass-and-straightedge grid, symmetry tools and SVG export.", "repo": null }, "author": { "name": "@fredini", "handle": "fredini", "url": "https://x.com/fredini", "platform": "X" }, "sources": [ { "url": "https://x.com/fredini/status/2090459022843379980", "platform": "X", "posted_at": "2026-08-20", "engagement": "3 likes", "engagement_n": 3 } ], "posted_at": "2026-08-20", "added_at": "2026-10-15" }, { "id": "update-client-records-from-one-telegram-message-but-only-when-confident", "url": "https://agentpicklist.com/use-cases/update-client-records-from-one-telegram-message-but-only-when-confident/", "title": "Update client records from one Telegram message, but only when confident", "summary": "An automation builder made an n8n flow that updates Notion and Sheets from a Telegram message and refuses below 80% match confidence.", "what_they_did": "@iamvaarr has two clients named Ahmed, so below 80% match confidence the workflow asks him instead of writing.", "outcome": "His point: a wrong write is silent, while a question takes 4 seconds.", "worth_knowing": "Confidence thresholds prevent silent data corruption.", "category": "ops", "agents": [ "n8n" ], "tools": [ "Telegram", "Notion", "Google Sheets" ], "roles": [ "Operations", "Freelancer" ], "industries": [ "Services" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "n8n", "Telegram bot", "Notion", "Google Sheets" ], "steps": [ "Write down the manual process step by step.", "Automate the deterministic parts with plain workflow steps or scripts.", "Use the model only where judgment is needed, with confidence checks.", "Add alerts and a human fallback." ], "prompt": "Match this message to a client. If confidence is below 0.8, ask me which client instead of writing anything.", "repo": null }, "author": { "name": "@iamvaarr", "handle": "iamvaarr", "url": "https://x.com/iamvaarr", "platform": "X" }, "sources": [ { "url": "https://x.com/iamvaarr/status/2090085320322888179", "platform": "X", "posted_at": "2026-08-19", "engagement": "", "engagement_n": 0 } ], "posted_at": "2026-08-19", "added_at": "2026-10-15" }, { "id": "do-customer-support-from-your-phone-with-the-claude-app", "url": "https://agentpicklist.com/use-cases/do-customer-support-from-your-phone-with-the-claude-app/", "title": "Do customer support from your phone with the Claude app", "summary": "An indie app founder now handles support, feature triage and bug fixes from anywhere with the Claude app.", "what_they_did": "@jackfriks uses the Claude app on phone or laptop for customer support.", "outcome": "He says it goes beyond basic support to triaged feature requests and bug fixes, better than he managed before.", "worth_knowing": "Keep a human check on refunds and account actions.", "category": "support", "agents": [ "Claude" ], "tools": [], "roles": [ "Indie founder" ], "industries": [ "Consumer apps" ], "difficulty": "beginner", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude app", "Support inbox access", "Repo access" ], "steps": [ "Collect the incoming channel (inbox, form, tickets) into one workflow.", "Have the model classify and extract fields, returning structured JSON.", "Route or draft replies; keep a human approval step for anything customer-facing.", "Log results and review misclassifications weekly." ], "prompt": "Read today’s support emails, group them into bugs, feature requests and how-to questions, draft replies, and open a fix for the top bug.", "repo": null }, "author": { "name": "@jackfriks", "handle": "jackfriks", "url": "https://x.com/jackfriks", "platform": "X" }, "sources": [ { "url": "https://x.com/jackfriks/status/2089725158005121225", "platform": "X", "posted_at": "2026-08-18", "engagement": "104 likes", "engagement_n": 104 } ], "posted_at": "2026-08-18", "added_at": "2026-10-15" }, { "id": "build-a-personal-cfo-dashboard-that-catches-money-leaks", "url": "https://agentpicklist.com/use-cases/build-a-personal-cfo-dashboard-that-catches-money-leaks/", "title": "Build a personal CFO dashboard that catches money leaks", "summary": "A builder used Claude Code to build a financial command center and says it saved him over $500 last month.", "what_they_did": "@AugmentStartups built a custom AI CFO dashboard with Claude Code.", "outcome": "Reports saving over $500 in a month.", "worth_knowing": "Self-reported savings.", "category": "finance", "agents": [ "Claude Code" ], "tools": [], "roles": [ "Founder" ], "industries": [ "Finance" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": "Sourced: over $500 saved in a month (self-reported)", "recipe": { "stack": [ "Claude Code", "Bank/card exports", "A dashboard framework" ], "steps": [ "Connect or export your financial data.", "Have the agent categorize and summarize it.", "Check the numbers against your statements.", "Schedule recurring reports." ], "prompt": "Load my last 3 months of transactions, find recurring charges and duplicates, and show a dashboard of potential monthly savings.", "repo": null }, "author": { "name": "@AugmentStartups", "handle": "AugmentStartups", "url": "https://x.com/AugmentStartups", "platform": "X" }, "sources": [ { "url": "https://x.com/AugmentStartups/status/2089261302510182553", "platform": "X", "posted_at": "2026-08-17", "engagement": "2 likes", "engagement_n": 2 } ], "posted_at": "2026-08-17", "added_at": "2026-10-15" }, { "id": "sell-an-ai-support-agent-to-shopify-stores", "url": "https://agentpicklist.com/use-cases/sell-an-ai-support-agent-to-shopify-stores/", "title": "Sell an AI support agent to Shopify stores", "summary": "A founder built an AI support agent for Shopify stores and got a first customer on a $99/month plan.", "what_they_did": "@oderajoseph built Arbyn. A user upgraded from the free plan after a few days to the $99/month Arbyn Agent plan.", "outcome": "First paying customer at $99/month.", "worth_knowing": "Very early: one customer.", "category": "ecommerce", "agents": [], "tools": [ "Shopify" ], "roles": [ "Founder" ], "industries": [ "E-commerce" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": { "usd_per_month": 99, "text": "$99/mo (first customer, self-reported)", "bucket": "under1k", "source_stated": true, "verified": false }, "result": null, "recipe": { "stack": [ "LLM", "Shopify API", "Help desk integration" ], "steps": [ "Connect store data (orders, products, policies).", "Have the agent answer or act within written policy.", "Hand off refunds and edge cases to a human.", "Review transcripts weekly." ], "prompt": "Answer this shopper using only the store’s policies and order data. If a refund is requested, collect details and hand off to a human.", "repo": null }, "author": { "name": "@oderajoseph", "handle": "oderajoseph", "url": "https://x.com/oderajoseph", "platform": "X" }, "sources": [ { "url": "https://x.com/oderajoseph/status/2089022188531982556", "platform": "X", "posted_at": "2026-08-16", "engagement": "10 likes", "engagement_n": 10 } ], "posted_at": "2026-08-16", "added_at": "2026-10-15" }, { "id": "automate-your-whole-job-search-then-choose-to-apply-by-hand", "url": "https://agentpicklist.com/use-cases/automate-your-whole-job-search-then-choose-to-apply-by-hand/", "title": "Automate your whole job search, then choose to apply by hand", "summary": "A builder made an n8n job search pipeline with contact lookup and tailored resumes, then kept applying manually.", "what_they_did": "@nethxkensoriano built job discovery, hiring-contact lookup, tailored applications and custom resumes in n8n with free tools.", "outcome": "He built it to learn n8n and kept applying manually anyway.", "worth_knowing": "Personal applications may still land better.", "category": "recruiting", "agents": [ "n8n" ], "tools": [], "roles": [ "Job seeker" ], "industries": [ "Recruiting" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "n8n", "Job board feeds", "LLM" ], "steps": [ "Define your criteria.", "Let the agent search, score and draft.", "Review everything before it is sent.", "Track outcomes." ], "prompt": "Find 20 roles matching [criteria], find a hiring contact for each, and draft a 3-sentence personal note. Do not send.", "repo": null }, "author": { "name": "@nethxkensoriano", "handle": "nethxkensoriano", "url": "https://x.com/nethxkensoriano", "platform": "X" }, "sources": [ { "url": "https://x.com/nethxkensoriano/status/2086665662030815541", "platform": "X", "posted_at": "2026-08-10", "engagement": "2 likes", "engagement_n": 2 } ], "posted_at": "2026-08-10", "added_at": "2026-10-15" }, { "id": "rebuild-a-classic-sim-game-with-a-new-twist-using-claude", "url": "https://agentpicklist.com/use-cases/rebuild-a-classic-sim-game-with-a-new-twist-using-claude/", "title": "Rebuild a classic sim game with a new twist using Claude", "summary": "A developer who has been rebuilding old games with Claude made a SimTower-like game whose economy runs on democratic socialism.", "what_they_did": "geekamongus has been rebuilding old games with Claude. While making a modern SimTower he made you an elected steward accountable to an Assembly instead of a tycoon.", "outcome": "Playable in the browser.", "worth_knowing": "Reusing a known game loop lets you focus on the twist.", "category": "gaming", "agents": [ "Claude" ], "tools": [ "Web" ], "roles": [ "Hobbyist", "Game designer" ], "industries": [ "Games" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude", "A browser game framework" ], "steps": [ "Describe the game or mod you want.", "Build it with your agent in small playable steps.", "Playtest after each step.", "Share or publish, respecting copyright." ], "prompt": "Build a browser tower-management game like [classic], but replace profit with resident satisfaction votes. Start with one floor type and a vote every 30 days.", "repo": null }, "author": { "name": "geekamongus", "handle": "geekamongus", "url": "https://news.ycombinator.com/user?id=geekamongus", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49195297", "platform": "Hacker News", "posted_at": "2026-08-06", "engagement": "41 points", "engagement_n": 41 } ], "posted_at": "2026-08-06", "added_at": "2026-10-15" }, { "id": "pause-your-music-when-claude-code-needs-you", "url": "https://agentpicklist.com/use-cases/pause-your-music-when-claude-code-needs-you/", "title": "Pause your music when Claude Code needs you", "summary": "A developer made Spotify pause whenever Claude Code needs input and resume when you answer.", "what_they_did": "@ironshay kept missing finished tasks because the music kept playing.", "outcome": "Released as an npm package.", "worth_knowing": "A small quality-of-life hack.", "category": "coding", "agents": [ "Claude Code" ], "tools": [ "Spotify" ], "roles": [ "Developer" ], "industries": [ "Software" ], "difficulty": "beginner", "time_to_build": { "text": "est. 2 minutes", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "claude-needs-you (npm)", "Spotify" ], "steps": [ "Open the repo in your coding agent and describe the goal and constraints.", "Ask for a plan first, then let it implement in small, reviewable steps.", "Run the tests or the app after each step and feed errors back.", "Review the diff before merging." ], "prompt": "No prompt needed: install with npm i -g claude-needs-you && cny init", "repo": null }, "author": { "name": "@ironshay", "handle": "ironshay", "url": "https://x.com/ironshay", "platform": "X" }, "sources": [ { "url": "https://x.com/ironshay/status/2084879702800371807", "platform": "X", "posted_at": "2026-08-05", "engagement": "2 likes", "engagement_n": 2 } ], "posted_at": "2026-08-05", "added_at": "2026-10-15" }, { "id": "vibe-code-an-app-to-8k-mrr-without-a-coding-background", "url": "https://agentpicklist.com/use-cases/vibe-code-an-app-to-8k-mrr-without-a-coding-background/", "title": "Vibe code a paid app to $8k MRR without a coding background", "summary": "A non-developer founder building apps with Claude Code reports reaching $8k in monthly recurring revenue after three months without sales.", "what_they_did": "@balubabear1312 switched from a previous business to building apps with Claude Code. He says he had zero success for about three months, made his first sale about 1.5 months before posting, and uses Apple Search Ads.", "outcome": "He reports $8k MRR, an $800 revenue day on $125 of Apple Search Ads spend, and says he wouldn't be surprised if MRR drops again.", "worth_knowing": "Self-reported and unverified. He did not name the app in this post. He credits paid acquisition (Apple Search Ads) rather than the code itself.", "category": "business", "agents": [ "Claude Code" ], "tools": [ "Apple Search Ads", "iOS" ], "roles": [ "Solo founder", "Non-developer" ], "industries": [ "Consumer apps" ], "difficulty": "intermediate", "time_to_build": { "text": "~4.5 months to traction (sourced)", "sourced": true }, "monthly_cost": { "text": "Sourced: $125 ad spend on the $800 day; tool costs not reported", "sourced": true }, "revenue": { "usd_per_month": 8000, "text": "$8k MRR (self-reported)", "bucket": "1k", "source_stated": true, "verified": false }, "result": null, "recipe": { "stack": [ "Claude Code", "Xcode / App Store", "Apple Search Ads", "RevenueCat or StoreKit for subscriptions" ], "steps": [ "Pick a niche with existing paid apps and clear search demand.", "Have Claude Code build a focused v1 with a subscription paywall.", "Start Apple Search Ads on a small daily budget on exact-match keywords.", "Track revenue per ad dollar and iterate the paywall and screenshots." ], "prompt": "Build a SwiftUI iOS app for [niche] with one core feature, onboarding in 3 screens, and a StoreKit 2 subscription paywall (weekly + yearly). Include analytics events for onboarding steps and paywall views.", "repo": null }, "author": { "name": "@balubabear1312", "handle": "balubabear1312", "platform": "X", "url": "https://x.com/balubabear1312" }, "sources": [ { "url": "https://x.com/balubabear1312/status/2082851165536952503", "platform": "X", "posted_at": "2026-07-30", "engagement_n": 0 } ], "posted_at": "2026-07-30", "added_at": "2026-10-15" }, { "id": "turn-your-ai-agent-into-a-bookkeeper-for-a-small-business", "url": "https://agentpicklist.com/use-cases/turn-your-ai-agent-into-a-bookkeeper-for-a-small-business/", "title": "Turn your AI agent into a bookkeeper for a small business", "summary": "A small business owner made a CLI that lets an AI agent keep simple books.", "what_they_did": "kvisner owns a small tech business with 10-30 transactions a month and built Magpie, a CLI that turns an agent into a bookkeeper.", "outcome": "Released open source.", "worth_knowing": "Have an accountant review year-end figures.", "category": "finance", "agents": [], "tools": [ "CLI" ], "roles": [ "Small business owner" ], "industries": [ "Finance" ], "difficulty": "intermediate", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Magpie CLI", "Any coding agent", "Bank CSV exports" ], "steps": [ "Connect or export your financial data.", "Have the agent categorize and summarize it.", "Check the numbers against your statements.", "Schedule recurring reports." ], "prompt": "Import this month’s bank CSV, categorize each transaction, flag anything unusual, and produce a profit and loss summary.", "repo": "https://github.com/kyle-visner/magpie" }, "author": { "name": "kvisner", "handle": "kvisner", "url": "https://news.ycombinator.com/user?id=kvisner", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=49091292", "platform": "Hacker News", "posted_at": "2026-07-28", "engagement": "8 points", "engagement_n": 8 } ], "posted_at": "2026-07-28", "added_at": "2026-10-15" }, { "id": "direct-a-10-minute-ai-movie-with-claude-code-as-the-director", "url": "https://agentpicklist.com/use-cases/direct-a-10-minute-ai-movie-with-claude-code-as-the-director/", "title": "Direct a 10-minute AI movie with Claude Code as the director", "summary": "A creator built a Claude Code pipeline that glues video, image and voice models together to make 10-minute films.", "what_they_did": "dawndrain uses Claude as director, gluing Seedance (video), Nano Banana (images) and ElevenLabs (voice). The repo has a Markdown playbook and a worked example.", "outcome": "He shared a first pass and the playbook for others.", "worth_knowing": "Generation costs add up; prototype short scenes first.", "category": "content", "agents": [ "Claude Code" ], "tools": [ "Seedance", "Nano Banana", "ElevenLabs" ], "roles": [ "Creator", "Filmmaker" ], "industries": [ "Media", "Entertainment" ], "difficulty": "advanced", "time_to_build": { "text": "Not reported", "sourced": false }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": null, "result": null, "recipe": { "stack": [ "Claude Code", "Video, image and voice APIs", "movie-gen playbook" ], "steps": [ "Pick a repeatable format.", "Let the agent generate drafts or assets from your inputs.", "Edit and check facts and rights before publishing.", "Reuse the template for the next piece." ], "prompt": "Write a 10-scene script for a 10-minute short about [idea], then a shot list with image prompts, video prompts and voice lines for each scene.", "repo": "https://github.com/dawndrain/movie-gen" }, "author": { "name": "dawndrain", "handle": "dawndrain", "url": "https://news.ycombinator.com/user?id=dawndrain", "platform": "Hacker News" }, "sources": [ { "url": "https://news.ycombinator.com/item?id=48978961", "platform": "Hacker News", "posted_at": "2026-07-20", "engagement": "21 points", "engagement_n": 21 } ], "posted_at": "2026-07-20", "added_at": "2026-10-15" }, { "id": "run-a-solo-automation-agency-with-claude-code", "url": "https://agentpicklist.com/use-cases/run-a-solo-automation-agency-with-claude-code/", "title": "Run a solo automation agency on Claude Code and reusable skills", "summary": "A solo operator describes building a $40k/mo services business with Claude Code, a skills folder, and tiered offers from websites to monthly systems.", "what_they_did": "@DeRonin_ shares a six-month roadmap: pick one niche, set up Claude Code with a company folder and GitHub backup, build demos for real businesses before you have clients, and turn anything you build twice into a reusable skill. Offers are tiered: $500 sites, $5k automations, and $2-5k/mo systems.", "outcome": "He says month 6 looks like a $40k/mo solo business.", "worth_knowing": "Self-reported in a promotional thread; the breakdown of revenue by tier isn't given. A services business, so revenue scales with your time and client count.", "category": "business", "agents": [ "Claude Code" ], "tools": [ "GitHub", "Claude Code skills" ], "roles": [ "Freelancer", "Agency owner" ], "industries": [ "Services", "Small business" ], "difficulty": "intermediate", "time_to_build": { "text": "~6 months (sourced)", "sourced": true }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": { "usd_per_month": 40000, "text": "$40k/mo (self-reported)", "bucket": "10k", "source_stated": true, "verified": false }, "result": null, "recipe": { "stack": [ "Claude Code", "A company folder with reusable skills", "GitHub", "A portfolio site" ], "steps": [ "Pick one niche and list the 4 automations every business in it needs.", "Build 3-5 unpaid demos for real businesses in that niche.", "Package offers in tiers: one-off site, one-off automation, monthly retainer.", "Turn every repeated build into a skill so the next one ships in hours." ], "prompt": "Act as my operations partner for a [niche] automation agency. From this client brief, pick the matching skill in /skills, adapt it, and produce: the build, a 1-page handover doc, and a monthly retainer proposal.", "repo": null }, "author": { "name": "@DeRonin_", "handle": "DeRonin_", "platform": "X", "url": "https://x.com/DeRonin_" }, "sources": [ { "url": "https://x.com/DeRonin_/status/2077387166389305576", "platform": "X", "posted_at": "2026-07-15", "engagement": "205 likes", "engagement_n": 205 } ], "posted_at": "2026-07-15", "added_at": "2026-10-15" }, { "id": "sell-a-linkedin-prospecting-agent-as-saas", "url": "https://agentpicklist.com/use-cases/sell-a-linkedin-prospecting-agent-as-saas/", "title": "Sell a LinkedIn prospecting AI agent as a SaaS product", "summary": "A founder built Gojiberry, a SaaS around a LinkedIn AI agent, and per a podcast host grew it from $50K to $300K MRR.", "what_they_did": "Florian Darroman says he interviewed @romanbuildsaas twice. The first time, Gojiberry had gone from $0 to $50K MRR in 180 days; by the second, it had hit $300K MRR. The episode covers distribution channels and Y Combinator lessons.", "outcome": "Reported $300K MRR (about $3.5M a year).", "worth_knowing": "Figures come from the podcast host's post, not audited data. LinkedIn automation must respect LinkedIn's user agreement and anti-spam laws.", "category": "business", "agents": [], "tools": [ "LinkedIn" ], "roles": [ "Founder", "Sales" ], "industries": [ "B2B SaaS", "Sales" ], "difficulty": "advanced", "time_to_build": { "text": "$50K MRR in 180 days (sourced)", "sourced": true }, "monthly_cost": { "text": "Not reported", "sourced": false }, "revenue": { "usd_per_month": 300000, "text": "$300K MRR (reported on a podcast)", "bucket": "100k", "source_stated": true, "verified": false }, "result": null, "recipe": { "stack": [ "An LLM agent framework", "CRM integration", "Compliant data sources" ], "steps": [ "Pick one buyer persona and one trigger signal (e.g. job change).", "Have the agent find matches and draft personalized first messages for human approval.", "Charge per seat and measure meetings booked.", "Grow through the channels your buyers already use." ], "prompt": "From this list of [persona] accounts and their recent posts, pick the 10 with the strongest buying signal for [product], and draft a 2-sentence personal opener for each that references the signal. Do not send anything.", "repo": null }, "author": { "name": "@romanbuildsaas (via Florian Darroman)", "handle": "floriandarroman", "platform": "X", "url": "https://x.com/floriandarroman" }, "sources": [ { "url": "https://x.com/floriandarroman/status/2077365562376663427", "platform": "X", "posted_at": "2026-07-15", "engagement_n": 0 } ], "posted_at": "2026-07-15", "added_at": "2026-10-15" } ] }