Strava

What is Strava?

An MCP server that provides 11 tools to interact with a user's Strava account, organized into activities, athlete profile, segments, routes, and analysis. It also includes a resource for athlete context and prompt templates for common queries, enabling access through Claude, ChatGPT, and other LLMs.

How to use Strava?

Install using Python 3.11+ with uv or Docker, then run the interactive authentication wizard (strava-mcp-auth). Add the server to your Claude Desktop configuration using uv run or Docker commands, and then ask natural language questions about your activities, segments, routes, or training analysis.

Key features of Strava

• 11 tools across 5 categories (activities, athlete, segments, routes, analysis)
• 1 MCP resource for ongoing athlete profile context
• 6 built-in prompt templates for common queries
• Supports both stdio (single-user) and HTTP (multi-user) modes
• OAuth-based authentication with automatic token refresh
• Cursor-based pagination with sensible defaults (10 items for most lists)

Use cases of Strava

• Analyze recent training over a specified time period
• Explore and track performance on specific segments
• Compare multiple runs or races side-by-side
• Export routes to GPX or TCX format
• Find activities similar to a reference workout for progress tracking

FAQ from Strava

What are the prerequisites for running Strava?

You need Python 3.11+ and uv, or Docker. A Strava API application must be created at strava.com with the callback domain set to localhost.

How does authentication work?

For stdio mode, a one-time OAuth flow through the browser saves tokens to a .env file with auto-refresh. For HTTP mode, per-user OAuth flow stores tokens in a session store (in-memory or DynamoDB) with a configurable expiration (default 12 hours).

What data can I access with this server?

The server can read your athlete profile and zones, all activity data, activity summaries, and star/unstar segments. The required OAuth scopes include profile:read_all, activity:read_all, activity:read, and profile:write.

Can I use this server with ChatGPT?

Yes, by running the server in HTTP mode and exposing it via ngrok or a public URL. Set STRAVA_MCP_BASE_URL to the ngrok URL and use that URL (with /mcp path) as your MCP server endpoint in ChatGPT.

What transport modes does the server support?

Two modes: stdio (default, single-user via standard input/output) and HTTP (streamable HTTP, multi-user with separate sessions). Select with the --transport flag (e.g., --transport http).