Access Strava Athlete Activity Data

Strava API MCP Server is a Model Context Protocol server that gives an AI assistant access to the Strava API so it can query the authenticated athlete's activities - recent runs, rides, and other workouts - directly from a conversation.

Use this when someone wants to ask an AI assistant natural-language questions about their Strava activity history, such as their recent activities, activities from a specific date range, their longest run in the past month, or details of their latest cycling session, instead of opening the Strava app or website. Requires the user's own Strava API application credentials and a refresh token, so it only works for the athlete who owns those credentials.

Four tools: get_activities(limit) returns the authenticated athlete's recent activities up to a limit (default 10); get_activities_by_date_range(start_date, end_date, limit) returns activities between a start and end date in ISO format (YYYY-MM-DD), defaulting to a limit of 30; get_activity_by_id(activity_id) returns detailed information about one specific activity; and get_recent_activities(days, limit) returns activities from the past N days (default 7). Every activity is returned with consistent field names and units: name, sport_type, start_date (ISO 8601), distance_metres, elapsed_time_seconds, moving_time_seconds, average_speed_mps, max_speed_mps, total_elevation_gain_metres, elev_high_metres, elev_low_metres, calories, and start/end latlng coordinates. The server also returns human-readable error messages for invalid date formats, API authentication errors, and network connectivity problems, rather than raw API error payloads. Typical example queries once connected include "What are my recent activities?", "Show me my activities from last week", "What was my longest run in the past month?", and "Get details about my latest cycling activity."

Requires Python 3.10. First create a Strava API application at strava.com/settings/api to get a Client ID and Client Secret, setting the Authorization Callback Domain to localhost. Then generate a refresh token by running the included script:

python get_strava_token.py

which walks through authorization and saves the resulting tokens to a .env file. Add the server to Claude Desktop's claude_desktop_config.json:

{
    "mcpServers": {
        "strava": {
            "command": "uvx",
            "args": ["strava-mcp-server"],
            "env": {
                "STRAVA_CLIENT_ID": "YOUR_CLIENT_ID",
                "STRAVA_CLIENT_SECRET": "YOUR_CLIENT_SECRET",
                "STRAVA_REFRESH_TOKEN": "YOUR_REFRESH_TOKEN"
            }
        }
    }
}

For Claude Web, run the server locally and connect it via the MCP extension.

Athletes and developers who want an AI assistant to answer natural-language questions about their own Strava training history - runs, rides, elevation, pace, calories - without switching to the Strava app, released under the MIT license.