Documentation

Use with AI

Let AI tools like Cursor and Claude Code search Guardy docs to instrument your agents automatically.

Guardy provides an MCP (Model Context Protocol) server that lets AI coding tools search our documentation directly. Ask your AI to "instrument my agent with Guardy" and it will pull the right docs, code examples, and API references automatically.

Install with Claude Code

Run this in your terminal:

claude mcp add --transport http Guardy https://tryguardy.com/docs/mcp

That's it. The SearchGuardy tool is now available in all your Claude Code sessions.

Install with Cursor

1. Open Cursor Settings (Cmd + , / Ctrl + ,)
2. Go to the MCP section
3. Click Add new MCP server
4. Set the type to SSE and the URL to:

https://tryguardy.com/docs/mcp

Or add it directly to .cursor/mcp.json in your project:

{
  "mcpServers": {
    "guardy": {
      "type": "sse",
      "url": "https://tryguardy.com/docs/mcp"
    }
  }
}

What You Can Do

Once connected, just ask your AI assistant naturally:

• *"Instrument my agent with Guardy"*
• *"Add Guardy tracking to my LangChain agent"*
• *"How do I track tool calls with the Python SDK?"*
• *"Set up Guardy with the Vercel AI SDK"*

The AI will search our docs and write the integration code for you.

MCP Prompt Tips

For Claude Agent SDK integrations, include your language and runtime behavior in the prompt so the assistant returns the correct pattern:

• *"Set up Claude Code Python tracking with wrap_claude_agent (for query()) or wrap_claude_client (for ClaudeSDKClient), and make sure sessions finalize with tokens/cost."*
• *"For Claude Code TypeScript, ensure early stream exits are handled so sessions don't stay running."*
• *"Show me the error-handling pattern for Claude Code wrappers, including generator-close behavior."*
• *"I'm using ClaudeSDKClient for multi-turn conversations — show me how to track each turn as a Guardy session with wrap_claude_client."*

How It Works

The MCP server exposes a single SearchGuardy tool. When your AI assistant needs information about Guardy, it calls this tool with a search query and gets back relevant documentation with code examples, API references, and direct links.

No API key is needed — the docs MCP is public and read-only.