A Model Context Protocol (MCP) server for Strava integration. Access your activities, athlete stats, segments, and routes through Claude, ChatGPT, and other LLMs.

This MCP server provides 11 tools to interact with your Strava account, organized into 5 categories:

• Activities (2 tools) - Query and analyze your Strava activities
• Athlete (1 tool) - Access profile, statistics, and training zones
• Segments (3 tools) - Explore and track segment efforts
• Routes (2 tools) - Manage and export routes
• Analysis (3 tools) - Training insights, comparison, and similarity search

Additionally, the server provides:

• 1 MCP Resource - Athlete profile with stats and zones for ongoing context
• 5 MCP Prompts - Templates for common queries (training analysis, segment performance, activity analysis, run comparison, training summary)

Deployment Options:

• Stdio Mode: Single-user local deployment (e.g., Claude Desktop, Cursor)
• HTTP Mode: Multi-user deployment (e.g., ChatGPT, remote MCP deployment)

• Python 3.11+ and uv, OR
• Docker

Before installation, you need to create a Strava API application:

1. Go to https://www.strava.com/settings/api
2. Create a new application with the following settings:
   • Application Name: Your choice (e.g., "My Strava MCP")
   • Category: Your choice (e.g., "Data Importer")
   • Club: Leave blank (optional)
   • Website: Can be anything (e.g., http://localhost)
   • Authorization Callback Domain: Must be localhost
3. After creation, note your Client ID and Client Secret
4. You'll use these credentials during the setup process below

1. OAuth Flow - One-time authorization through browser
2. Token Storage - OAuth tokens saved to .env file
3. Auto-Refresh - Tokens automatically refreshed when expired
4. Persistence - Subsequent runs reuse stored tokens

1. OAuth Flow - Per-user authorization on first connection (MCP OAuth → Strava OAuth)
2. Token Storage - OAuth tokens stored in session store (in-memory or DynamoDB)
3. Auto-Refresh - Tokens refreshed automatically per-session
4. Persistence - Sessions expire after 12 hours (configurable)

# Install dependencies
cd strava-mcp
uv sync

Then configure credentials:

uv run strava-mcp-auth

This interactive wizard will guide you through configuring authentication for stdio or http transport modes.

# Pull the image
docker pull ghcr.io/eddmann/strava-mcp:latest

Then configure credentials:

# Create the env file first (Docker will create it as a directory if it doesn't exist)
touch strava-mcp.env

# Run the setup script
docker run -it --rm \
  -v "/ABSOLUTE/PATH/TO/strava-mcp.env:/app/.env" \
  --entrypoint= \
  ghcr.io/eddmann/strava-mcp:latest \
  python -m strava_mcp.scripts.setup_auth

This interactive wizard will guide you through configuring authentication for stdio or http transport modes.

The authentication process requests these scopes:

• profile:read_all - Read athlete profile and zones
• activity:read_all - Read all activity data
• activity:read - Read activity summaries
• profile:write - Star/unstar segments

The server supports two transport modes selected via --transport flag (stdio is default, no flag needed):

Uses standard input/output for communication with a single pre-configured Strava account.

• Authentication: Pre-configured OAuth tokens in .env file
• Users: Single user per deployment
• Setup: Run strava-mcp-auth to authorize once
• Token Storage: Local .env file

Uses HTTP transport with per-user OAuth flow.

• Authentication: OAuth flow per user (MCP OAuth → Strava OAuth)
• Users: Multi-user support with separate sessions
• Setup: Environment-based configuration
• Token Storage: In-memory or DynamoDB session store

Add to your configuration file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "strava": {
      "command": "uv",
      "args": [
        "run",
        "--directory",
        "/ABSOLUTE/PATH/TO/strava-mcp",
        "strava-mcp",
        "--transport",
        "stdio"
      ]
    }
  }
}

{
  "mcpServers": {
    "strava": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-v",
        "/ABSOLUTE/PATH/TO/strava-mcp.env:/app/.env",
        "ghcr.io/eddmann/strava-mcp:latest",
        "--transport",
        "stdio"
      ]
    }
  }
}

Start the server in HTTP mode for deploying remote:

# Using UV
uv run --directory /ABSOLUTE/PATH/TO/strava-mcp strava-mcp --transport http

# Using Docker
docker run -p 8000:8000 --rm \
  -v "/ABSOLUTE/PATH/TO/strava-mcp.env:/app/.env" \
  ghcr.io/eddmann/strava-mcp:latest \
  --transport http

Environment variables can be configured using the strava-mcp-auth setup wizard (see Installation & Setup above).

To test ChatGPT integration locally:

1. Start the server in HTTP mode:

   uv run strava-mcp --transport http

2. In a separate terminal, expose via ngrok:

   ngrok http 8000

3. Update environment: Set STRAVA_MCP_BASE_URL to your ngrok URL:

   export STRAVA_MCP_BASE_URL=https://abc123.ngrok.io

   Or add to your .env file:

   STRAVA_MCP_BASE_URL=https://abc123.ngrok.io
   

4. Restart the server to pick up the new base URL

5. Configure ChatGPT: Use the ngrok URL (with /mcp path) as your MCP server endpoint

Ask Claude to interact with your Strava data using natural language. The server provides tools, a resource, and prompt templates to help you get started.

Use built-in prompt templates for common queries (available via prompt suggestions in Claude):

• analyze-recent-training - Analyze my training over the past 30 days
• segment-performance - Analyze my performance on a specific segment
• activity-deep-dive - Deep dive into a specific activity
• compare-recent-runs - Compare my recent runs to track progress
• training-summary - Show me a comprehensive training summary

"Show me my runs from the last 30 days"
"Get my last long run with lap splits and heart rate zones"
"Show me the comments and kudos on my half marathon race"
"Find all my marathon races from this year"
"Show me all 10km runs from the last 90 days"

"Show my athlete profile with year-to-date stats and training zones"

Note: The athlete profile resource (strava://athlete/profile) automatically provides ongoing context.

"Show me the Box Hill segment with my effort history"
"Find running segments near Hyde Park"
"Show the leaderboard for my favorite climb segment"

"List my routes"
"Export my Richmond Park route to GPX"

"Analyze my training over the past 30 days"
"Analyze my marathon race performance this year"
"Compare my last three 10K runs"
"Find runs similar to my last tempo workout"
"Show me all my 5k races from the last year"

Note: List-returning tools use cursor-based pagination with default limits (10 items for activities/segments/routes, 50 for leaderboards).

Resources provide ongoing context to the LLM without requiring explicit tool calls:

Prompt templates for common queries (accessible via prompt suggestion in Claude):

MIT License - see LICENSE file for details

This project is not affiliated with, endorsed by, or sponsored by Strava, Inc. All product names, logos, and brands are property of their respective owners.