Strava MCP Server

An open-source Model Context Protocol (MCP) server that connects AI agents to your Strava data.

Features

• Headless Adapter: No UI, just pure MCP tools and resources.
• Privacy First: Tokens stored locally (~/.config/strava-mcp/), never sent to 3rd parties.
• Resilient: Built-in rate limiting (handles 429s) and SQLite caching.
• Educational: Clean, readable Python 3.11+ code with type hints.

Prerequisites

• Python 3.11 or higher
• Strava Account (for API credentials)

Quick Start (Dry Run)

You can test the server immediately without Strava credentials using the mock data mode.

# Install dependencies
python3 -m pip install -e .

# Run the smoke test (verifies MCP protocol via stdio)
python3 scripts/smoke_test.py

Setup (Real Data)

1. Get Strava Credentials

1. Go to Strava API Settings.
2. Create an application.
   • Callback Domain: localhost

2. Authorization

Run the local helper script to perform the OAuth handshake. This will open your browser.

export STRAVA_CLIENT_ID=your_client_id
export STRAVA_CLIENT_SECRET=your_client_secret
python3 scripts/strava_auth.py

Tokens are saved to ~/.config/strava-mcp/credentials.json.

3. Usage with Claude Code

Configure claude to use this server.

File: ~/.config/claude/config.json (or equivalent MCP manager config)

{
  "mcpServers": {
    "strava": {
      "command": "python3",
      "args": ["-m", "strava_mcp.server"],
      "env": {
        "PYTHONPATH": "/path/to/strava-mcp/src" 
      }
    }
  }
}

Or run directly if installed in your environment:

{
  "mcpServers": {
    "strava": {
      "command": "uv",
      "args": ["run", "python", "-m", "strava_mcp.server"]
    }
  }
}

Tools Available

Tool	Description
get_athlete_profile	Returns your profile stats.
list_activities	Fetch recent activities (supports date filtering).
get_activity	Detailed view of a single activity.
weekly_summary	Aggregated stats (distance, elevation) for a given week.

Resources

• strava://athlete/profile: Direct read of profile.
• strava://activities/recent: Most recent 10 activities.

Architecture

• Tech Stack: Python, FastMCP, Httpx, SQLite, Pydantic.
• Caching: GET requests are cached in SQLite with a TTL (default 5 mins for lists).
• Rate Limiting: Respects X-RateLimit-* headers to prevent API bans.