AI
Data Analytics

AI Endurance MCP Server: Claude Integration and Analytics

Connect your AI Endurance platform to Claude via the MCP server to access workouts, performance analytics, and manage training plans conversationally.

View on GitHub
Quick Install
npx -y @ai-endurance/mcp

Overview

The AI Endurance MCP Server exposes your AI Endurance training data and analytics through the Model Context Protocol (MCP), enabling Claude and other MCP-compatible assistants to query and manipulate workouts, performance metrics, and training plans conversationally. It translates structured training objects (workouts, activities, zones, recovery metrics) into a form consumable by LLM-based assistants, and accepts commands to create or reschedule workouts so those changes propagate back to connected platforms (Garmin, TrainingPeaks, etc.).

For developers this server is useful when you want to add conversational access to endurance training data — for example, to let an assistant answer “How was my ride yesterday?” with power curve and stress scores, or to create a structured interval workout by voice. The server supports modern MCP clients using streamable HTTP transports (recommended) and also provides a legacy SSE transport configuration for older clients.

Features

  • Training plan management: view, edit, create and reschedule structured workouts (running, cycling, swimming)
  • Activity analysis: power curves, pace trends, normalized power, zone breakdowns, and time-series metrics
  • Performance predictions: ML-based race time predictions and fitness forecasts with confidence intervals
  • Recovery tracking: HRV, resting heart rate trends, and readiness/recovery scores
  • Zone management: view and update pace/power training zones
  • Workout scheduling: move workouts, set availability, and sync plan progress
  • Race goals: manage primary and secondary race targets and assess realism against current fitness
  • Multiple transport support: streamable HTTP (recommended) and SSE (legacy)

Installation / Configuration

Prerequisites:

  • An AI Endurance account
  • OAuth client credentials (authorization and token endpoints)
  • Claude Pro or another MCP 2025-06-18-compatible client

Quick Claude connector setup:

  1. In Claude, open Settings → Connectors → Add custom connector
  2. Provide the remote MCP Server URL and follow the OAuth flow

Example Claude configuration:

Name: AI Endurance
Remote MCP Server URL: https://aiendurance.com/mcp

Recommended MCP client configuration (streamable HTTP transport):

{
  "url": "https://aiendurance.com/mcp",
  "transport": {
    "type": "http"
  },
  "auth": {
    "type": "oauth",
    "authorizationUrl": "https://aiendurance.com/authorize/",
    "tokenUrl": "https://aiendurance.com/api/o/token/",
    "scopes": ["read", "write"]
  }
}

Legacy SSE transport:

{
  "url": "https://aiendurance.com/mcp",
  "transport": {
    "type": "sse"
  },
  "auth": {
    "type": "oauth",
    "authorizationUrl": "https://aiendurance.com/authorize/",
    "tokenUrl": "https://aiendurance.com/api/o/token/",
    "scopes": ["read", "write"]
  }
}

Available Resources

  • Repository and source: https://github.com/ai-endurance/mcp
  • OAuth endpoints:
    • Authorization: https://aiendurance.com/authorize/
    • Token: https://aiendurance.com/api/o/token/
  • Recommended scopes: read, write
  • Compatible clients: Claude Desktop, Cursor, Continue (VS Code), Cline, and any MCP 2025-06-18 client

Compatibility table:

FeatureSupported
Streamable HTTP transportYes (recommended)
SSE transportYes (legacy)
OAuth authenticationYes
Structured workouts (intervals, sets)Yes
Activity analytics (power curves, HRV)Yes

Use Cases

  1. Conversational analytics
    • Example: “How was my ride yesterday?”
      • Assistant returns power distribution, normalized power, External Stress Score, duration, and zone breakdown; can surface notable efforts or fatigue markers.
  2. Training plan editing
    • Example: “Move tomorrow’s threshold workout to Friday”
      • Server updates the workout date, recalculates plan progress, and can sync changes to connected services.
  3. Custom workout creation
    • Example: “Create a threshold run for tomorrow: 15min warmup, 3x8min threshold with 2min recoveries, 10min cooldown”
      • Server builds a structured workout with appropriate zones and publishes it to a linked device or platform.
  4. Recovery-informed recommendations
    • Example: “Am I recovered enough for today’s hard workout?”
      • Server returns recovery score, HRV and resting HR trends, and a recommendation (proceed / modify / rest) with rationale.
  5. Race planning and prediction
    • Example: “What’s my predicted half marathon time based on current fitness?”
      • Server returns an ML-based prediction, confidence interval, and suggested training adjustments to improve the target time.

Developer tips

  • Use the streamable HTTP transport when possible; it supports chunked/real-time responses and generally provides a better UX for LLM integrations.
  • Request only the scopes you need (read/write) and follow standard OAuth best practices for token storage and refresh.
  • For custom clients, surface structured workout objects (intervals, duration, intensity) to the assistant so it can create or edit workouts reliably.
  • Review the GitHub repo for code examples, issue tracker, and contribution guidelines: https://github.com/ai-endurance/mcp

This MCP server is focused on providing structured endurance data and actions to AI assistants — ideal for building conversational training assistants, analytics bots, or integrations that let athletes manage training hands-free.

Tags:data-analytics