v0.1.0 on PyPI

Give Claude access to your Strava data

An MCP server that turns your runs, rides, and swims into tools for AI-powered coaching, analysis, and conversation.

$ pip install strava-mcp-agent Copy
View Source
13
MCP Tools
16+
Strava Endpoints
0
Config Required
OAuth2
Auto Token Refresh
Tools
Everything in your Strava, accessible to Claude
13 tools covering activities, segments, routes, gear, and AI-ready coaching summaries.
get_athlete
Your profile — name, weight, FTP, bio, and stats
get_athlete_stats
All-time and recent totals for running, cycling, swimming
list_activities
Browse activities with date range and type filtering
get_activity
Full detail — splits, HR, cadence, weather, device
get_activity_zones
Heart-rate and power zone distribution
get_activity_laps
Per-lap pace, HR, and distance breakdown
get_activity_streams
Raw sensor data — GPS, HR, cadence, watts, altitude
get_starred_segments
All your starred segments at a glance
get_segment_efforts
Every effort on a specific segment
get_routes
Routes you've created on Strava
get_gear
Shoe and bike details with mileage tracking
get_clubs
Clubs you belong to
get_running_summary
AI-ready coaching summary — weekly mileage, pace trends, best efforts, HR stats
In Action
Talk to your training data
Just ask Claude naturally. It fetches the data, does the math, and gives you coaching insights.
You
How's my running been this month? Am I improving?
Claude
Looking at your data from the past 4 weeks, you've logged 12 runs totaling 87.3 km. Your average pace improved from 5:42/km to 5:21/km, and your resting HR on easy runs dropped by 4 bpm. Your best 5K effort this month was 24:18 — a 32-second PR. You're trending in the right direction. I'd suggest adding one more easy run per week to build your aerobic base further.
You
Which of my shoes has the most miles?
Claude
Your Nike Pegasus 40 has 847 km on them. Most running shoes are designed for 500–800 km — you might want to consider replacing them soon to avoid injury risk from worn-out cushioning.
You
Compare my heart rate zones on my last tempo run vs my long run
Claude
Your tempo run (Tuesday) spent 68% of the time in Zone 4 (threshold) with avg HR 172. Your long run (Saturday) was mostly Zone 2 (72%) with avg HR 142. That's a solid distribution — your easy runs are actually easy, and your hard runs are hard. Good polarization.
Quick Start
Running in 3 minutes
Install, configure your Strava credentials, and start talking to your data.

1. Install from PyPI

One command. No extra dependencies to manage.

2. Get Strava API credentials

Create an app at strava.com/settings/api and run python get_token.py

3. Add to Claude Desktop

Drop the config into your Claude Desktop settings and restart.

claude_desktop_config.json
{
  "mcpServers": {
    "strava": {
      "command": "strava-mcp",
      "env": {
        "STRAVA_CLIENT_ID": "your_id",
        "STRAVA_CLIENT_SECRET": "your_secret",
        "STRAVA_REFRESH_TOKEN": "your_token"
      }
    }
  }
}
Why
Built for athletes who use AI
⚡

Real-time data

OAuth2 token auto-refresh. Always current, never stale credentials.

🏃

Coaching-ready

The running summary tool computes weekly mileage, pace trends, best efforts, and HR stats in one call.

🔒

Secure by default

Credentials loaded from environment variables only. Never hardcoded, never logged.

🚀

Zero config server

Single command to run. Works with Claude Desktop, Claude Code, or any MCP client.

📈

Raw sensor streams

GPS, heart rate, cadence, power, altitude — every data point your device recorded.

🛠

Full Strava API

Activities, segments, routes, gear, clubs — if Strava has it, Claude can access it.

Start training with AI

Install in seconds. Ask Claude about your runs in minutes.

$ pip install strava-mcp-agent Copy
View on GitHub