AI assistants

Two ways to bring an AI assistant in: hand it these docs, so it writes your integration, or give it your project's server API as tools, so it can look events up for you.

The docs

• /llms.txt lists every docs page with a one-line summary, plus the security page and the legal policies, in plain text.
• Every page is server rendered HTML with the code of every tab in it (fetch, Node SDK, Python, PHP, Go and the frameworks), so an assistant that fetches a page sees every language, not only the one shown first.
• Copy for LLM at the top of every page copies its text with every tab's code, each under its label.

MCP server

@zeno-id/mcp gives an AI assistant your project's server API as tools: get_event, search_events, get_visitor, get_fingerprints and get_related (related IDs), and delete_visitor and forget_external_id. The two that erase data are marked destructive, so clients ask before running them, and they only run with confirm: true in their arguments.

Whatever the assistant reads through these tools, events with their IPs, locations and your external IDs, goes to the assistant's AI provider. That's a new recipient of your users' personal data: check your agreement with that provider (data processing, where it's processed) and your privacy policy cover it before you point the assistant at production data.

It needs Node (for npx) and a secret key, which reads every event of the project; make a separate one for the assistant on the API keys page, so you can revoke it on its own. With Claude Code:

claude mcp add zeno --env FP_SECRET_KEY=sk_... --env FP_ENDPOINT=https://api.zeno.id -- npx -y @zeno-id/mcp

Other clients take the same command, npx -y @zeno-id/mcp, with FP_SECRET_KEY and FP_ENDPOINT (the server API's origin, https://api.zeno.id) in its environment; both are required. It holds your secret key, so run it on your own machine or server.