Stop being your team's lookup service.
An MCP server that lets the company's internal AI assistant answer stakeholders' questions about what my team is working on — served straight from my laptop through a free tunnel.
- Built with
- Python · MCP
- Served over
- A free ngrok tunnel
- Access
- Read-only, curated pages only
- Status
- Proof of concept
Stakeholders ask the same question all week: what is your team actually working on? The assistant can reach the tools the platform team connected, but it cannot see my team's context — so every one of those questions came to me.
The edge of the platform is not the edge of what you can build. MCP is the seam. Anything I can reach, the assistant can reach — so I put a small server in front of my team's pages and let people ask it instead of me.
No hosting and no deployment. The server runs on my laptop, and a free ngrok tunnel gives it a public address that the assistant can call as a connector.
It only reads
There is no write path in the code. The worst case is a wrong answer, not a changed system.
It only sees the curated half
My meeting notes, drafts and personal pages sit in the same wiki. The server can never reach them — that part is decided in code, not in a setting.
It answers in structure
Tools return JSON rather than a wall of text, so the assistant gets facts it can use instead of prose it has to interpret.
- A PM can extend the platform. The assistant your company gives you is not the limit of what it can answer.
- The bar is lower than it looks. A Python file, a free tunnel, and a connector setting. No infrastructure, no ticket, no waiting.
- Stakeholders self-serve. The same question stops being a message to me and becomes something they can just ask.
- It is a proof of concept. It runs on my laptop behind a tunnel. It exists to show the shape of the thing — not to be the thing.