MCP server

Torii runs a hosted Model Context Protocol server that gives AI agents (Claude Code, Cursor, VS Code Copilot, and anything else that speaks MCP) direct access to these docs. With it connected, your AI agent looks up the real component props, hook signatures, and guides instead of guessing, so the code it writes matches the current SDK.

The server is public, read-only, and needs no API key.

Endpoint: https://mcp.toriiauth.eu/mcp (Streamable HTTP transport)

What it exposes

Two tools:

list_torii_docs — lists every documentation page (guides, components, hooks, customization, API reference) with an id, title, and tags.
get_torii_docs — returns the full Markdown for the page ids you ask for.

A typical AI agent lists the index, picks the pages relevant to your task, and reads them before writing code.

Add it to your editor

Run:

claude mcp add --transport http torii https://mcp.toriiauth.eu/mcp

Then start a session and ask it to “use the Torii docs to add sign-in”. Check it registered with claude mcp list.

Add the server to ~/.cursor/mcp.json (global) or .cursor/mcp.json (per project):

{
  "mcpServers": {
    "torii": {
      "url": "https://mcp.toriiauth.eu/mcp"
    }
  }
}

Add a workspace server in .vscode/mcp.json:

{
  "servers": {
    "torii": {
      "type": "http",
      "url": "https://mcp.toriiauth.eu/mcp"
    }
  }
}

Any client that speaks Streamable HTTP can point at the URL directly. For a client that only supports the older stdio transport, bridge it with mcp-remote:

{
  "mcpServers": {
    "torii": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.toriiauth.eu/mcp"]
    }
  }
}

Good prompts

Once connected, lead with intent and let the agent pull the docs it needs:

“Add Torii sign-in and sign-up to this React app using the Torii docs.”
“Wire up useAuth and gate this route on an authenticated session.”
“Set up first-party cookies for Torii before we ship to production.”