Let non-technical staff query sales data in plain English

An e-commerce company connected internal users’ agents to its data so they can ask for sales counts and charts themselves.

Difficulty
IntermediateOur rating
Time to build
Not reportedOur estimate
Monthly cost
Not reported
Revenue
Not reportedWe only show revenue the source states
Engagement
0 replies to HN commentComment in β€œAsk HN: Who is using MCP in production?” (198 points, 201 comments) Β· fetched Oct 15, 2026

What they did

NichoPaolucci set up a way for internal users to connect an agent to company data, replacing many hand-built reports.

How it went

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.

Builder recipe

Our own starting point, written from the source. Adapt it to your setup.

Stack

  • Read-only data access
  • MCP server
  • Claude or ChatGPT

Steps

  1. Connect store data (orders, products, policies).
  2. Have the agent answer or act within written policy.
  3. Hand off refunds and edge cases to a human.
  4. Review transcripts weekly.

Starter prompt

How many units of [product] sold each month in 2024? Make a chart and note any month that changed by more than 20%.

Tags

Read the original

Related use cases

The Daily 5.

Five real AI agent use cases each morning, each with a recipe and a source. Plus a Sunday roundup. Free.