AI agent in Slack for an e-commerce team

Context: built in-house for an e-commerce team · 2026 – present Stack: Claude (API), Python, MCP, REST APIs, Slack

This was built in-house, so it shows the design only. There is no code, no data and no company detail.

The problem

A commercial team runs on recurring data questions and routine operations: what sold, what's overstocked, whether a campaign is set up correctly. Each one meant waiting for an analyst, or someone doing it by hand.

What I built

An LLM agent the team talks to in Slack. It reads the data warehouse and the business tools, answers in the thread, and carries out routine tasks.

• Own agent harness in Python: a request queue, a router that keeps each conversation in its own session, and an explicit approval step before anything is written to a live system.
• Spend control: every request runs under a token budget, so a runaway conversation can't burn money.
• Tools over API and MCP: the data warehouse for numbers, the shop platform, and the team's existing business tools and workflows.
• Scheduled routines alongside the chat: recurring checks and reports that post to Slack on their own.

Result

• In daily production use by the team.
• Routine questions are answered in the thread, in minutes, instead of waiting for an analyst.
• Writes stay safe: nothing touches a live system without a person approving it.

What I'd bring to your team

The same pattern fits any team that asks its data the same questions every week: an agent in your chat tool, reading your data and tools, with approvals and a spend limit.