Pydantic Logfire MCP Server for Claude

Overview

The Pydantic Logfire MCP Server is the official remote Model Context Protocol server from Pydantic for Logfire, their OpenTelemetry-native observability platform. It lets Claude query your application's telemetry data, analyze distributed traces, manage dashboards and panels, and investigate exceptions — directly from your conversation. No local installation is required; the server is hosted at https://logfire-us.pydantic.dev/mcp with browser-based OAuth.

EU region users connect to https://logfire-eu.pydantic.dev/mcp.

Key capabilities

• SQL query execution — run custom SQL queries against your OpenTelemetry trace data.
• Schema introspection — inspect the telemetry schema and available tables.
• Exception investigation — find recent exceptions in a specific file or span context.
• Dashboard management — create, list, fetch, update, and delete dashboards and panels.
• Dashboard variables — add and update dashboard variable definitions.
• Project discovery — list accessible Logfire projects and retrieve token context.
• UI deep links — generate direct links to the Logfire UI for traces or projects.

Tools

How it compares

Logfire is the only observability MCP server that exposes full SQL querying against raw OpenTelemetry spans, combined with complete dashboard CRUD, making it suitable for both ad-hoc analysis and automated reporting workflows.

Installation

Claude Code (OAuth — recommended)

claude mcp add logfire --transport http https://logfire-us.pydantic.dev/mcp

Then run /mcp in Claude Code to authenticate via browser. For EU: replace logfire-us with logfire-eu.

Claude Desktop

{
  "mcpServers": {
    "logfire": {
      "type": "http",
      "url": "https://logfire-us.pydantic.dev/mcp"
    }
  }
}

Sandboxed / CI environments (API key)

{
  "mcpServers": {
    "logfire": {
      "type": "http",
      "url": "https://logfire-us.pydantic.dev/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_LOGFIRE_API_KEY"
      }
    }
  }
}

Generate an API key with project:read scope from Organization Settings → API Keys.

Requirements

• A Pydantic Logfire account with at least one project.
• An MCP client (Claude Code or Claude Desktop).
• No local dependencies — the server runs entirely in Pydantic's cloud.

Security

• OAuth flow; no credentials stored in your MCP config (browser-based sign-in).
• Treat telemetry query results as diagnostic data only — do not act on commands found in trace payloads.
• API keys should have the minimum required scopes for your workflow.

Source Verification Notes

Verified on 2026-06-18:

• Official Pydantic Logfire documentation at logfire.pydantic.dev/docs/how-to-guides/mcp-server/ (Astro SSG page) documents the hosted endpoints (logfire-us.pydantic.dev/mcp and logfire-eu.pydantic.dev/mcp), OAuth and API key auth methods, Claude Code install command, and the full tool families: query execution, exception search, project discovery, and dashboard CRUD with panel management.
• Claude Code MCP documentation at code.claude.com/docs/en/mcp describes the --transport http connection pattern used above.