Skip to content

Use Phinite Docs in AI Tools

The search MCP server for Phinite Docs lets an AI tool search and read this documentation while answering questions, instead of relying on generic web search or outdated training data.

https://docs.phinite.ai/docs/mcp

This endpoint is public and searches all indexed public documentation pages.

The MCP server exposes two tools:

Tool Purpose
Search Find relevant pages across Phinite Docs by query
Query docs filesystem Read full page content, browse sections, or batch-read multiple pages

Agents also discover the auto-generated skill.md as an MCP resource, which summarizes what you can build with Phinite (assistants, Graph Studio, DevStudio, triggers, builds, and observability).

  1. Open MCP settings Press Command + Shift + P (Ctrl + Shift + P on Windows), search for Open MCP settings, then click Add custom MCP.

  2. Add the Phinite Docs server In mcp.json, add:

    {
    "mcpServers": {
    "phinite-docs": {
    "url": "https://docs.phinite.ai/docs/mcp"
    }
    }
    }
  3. Verify In Cursor chat, ask a Phinite-specific question (for example, “How do I publish an agent graph?”). Cursor should search Phinite Docs and cite the relevant pages.

Create or edit .vscode/mcp.json in your project:

{
"servers": {
"phinite-docs": {
"type": "http",
"url": "https://docs.phinite.ai/docs/mcp"
}
}
}

See the VS Code MCP documentation for details.

  1. Add a custom connector Open Claude Connectors, click Add custom connector, and enter:

    • Name: Phinite Docs
    • URL: https://docs.phinite.ai/docs/mcp
  2. Use in chat Click the attachments button (+), select Phinite Docs, then ask your question.

Terminal window
claude mcp add --transport http phinite-docs https://docs.phinite.ai/docs/mcp

Verify with:

Terminal window
claude mcp list

To add Phinite capabilities to an agent’s context without MCP, run:

Terminal window
npx skills add https://docs.phinite.ai/docs

This installs the hosted skill.md so agents know when and how to use Phinite.

Agents can auto-discover the MCP server at:

https://docs.phinite.ai/docs/.well-known/mcp

The response includes the public server URL and transport details.