Skip to content

Use FOID docs with AI ​

Building with an AI assistant? Give it the FOID docs, so it writes code against the real SDKs and API instead of guessing.

There are three ways to do it:

Best forSetup
Copy page / Open in…A quick question about one pageNone — the menu at the top of every page
llms.txtPasting all the docs into a chat or a promptNone — a URL
MCP serverYour editor or agent searching the docs while it worksOne command

Copy a page ​

Every page has a Copy page button at the top right. Its menu has:

  • Copy page — the page as clean markdown (code samples included), ready to paste into any AI tool.
  • View as Markdown — the same text in your browser. Every page has one: add .md to the URL, e.g. /guide/quickstart.md.
  • Open in ChatGPT / Claude / Perplexity — opens a new chat that reads this page. When you run the docs on your own machine (localhost), those services can't reach it, so the page is copied to your clipboard and the chat asks you to paste it.

llms.txt ​

The docs follow the llms.txt convention:

URLWhat it is
/llms.txtAn index: every page with a one-line summary and a link to its markdown
/llms-full.txtEvery page in one file — paste it into a chat or a system prompt
/<page>.mdOne page as markdown

All three are generated from the same pages you're reading, every time the docs are built, so they never drift.

MCP server ​

FOID runs a Model Context Protocol server that lets AI tools search and read these docs on their own. It's read-only, needs no key, and only serves the public docs — nothing about any person or business.

URLhttps://api.foid.ng/mcp
Running FOID locallyhttp://localhost:8000/mcp (the backend; it reads the docs from the docs site, so run pnpm dev in docs/ too)
TransportStreamable HTTP

Tools

ToolWhat it does
search_foid_docsquery, optional limit (1–10) → the best-matching sections, each with an excerpt and a link
read_foid_docpath (e.g. /guide/continue-with-foid) → that page as markdown
list_foid_docsEvery page, grouped by section

Add it to your tool ​

bash
claude mcp add --transport http foid-docs https://api.foid.ng/mcp
json
{
  "mcpServers": {
    "foid-docs": { "url": "https://api.foid.ng/mcp" }
  }
}
json
{
  "servers": {
    "foid-docs": { "type": "http", "url": "https://api.foid.ng/mcp" }
  }
}
json
{
  "mcpServers": {
    "foid-docs": { "serverUrl": "https://api.foid.ng/mcp" }
  }
}
json
{
  "mcpServers": {
    "foid-docs": { "command": "npx", "args": ["-y", "mcp-remote", "https://api.foid.ng/mcp"] }
  }
}
  • Claude (desktop and claude.ai): Settings → Connectors → Add custom connector, name it FOID Docs, URL https://api.foid.ng/mcp.
  • ChatGPT: Settings → Apps & Connectors → Advanced settings → turn on Developer mode, then Create a connector with the URL https://api.foid.ng/mcp and no authentication.

Claude and ChatGPT connect from their own servers, so they need the public https:// URL. Editor tools like Claude Code, Cursor and VS Code run on your machine and can use http://localhost:8000/mcp.

Then just ask, for example: "Using the FOID docs, add Continue with FOID to my Next.js app with a Node backend."

Check it's working ​

bash
curl -s https://api.foid.ng/mcp -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"search_foid_docs","arguments":{"query":"verify webhook signature"}}}'

You should get back sections from the Webhooks guide.

Running FOID yourself

The MCP server is part of the backend: POST /mcp. It loads the docs from DOCS_SITE_URL/docs-index.json and caches them for 10 minutes. Turn it off with DOCS_MCP=false.

FOID — Fintech Open Identity. Test mode never touches real identity data.