Gym Coach Server

MCP server that turns any compatible LLM into a personal gym coach with access to your real workout data from Gym Tracker.

Built on the Model Context Protocol — works with Claude Desktop, Claude Code, Cursor, Continue, Cline, and any MCP-compatible client.

What it does

The server connects to your Gym Tracker Supabase database and exposes your training data through 5 tools and 1 resource:

Tools

Tool	Description	Parameters
get_workout_history	Recent workouts with exercises, weights, and reps	limit, from_date, to_date
get_exercise_progress	Progression history for a specific exercise (max weight, volume, trends)	exercise_name, limit
get_routines	All routines with exercises, sets, reps, and muscle groups	—
get_training_plan	Active training plan with progress and completion percentage	—
get_stats_summary	KPIs: total workouts, frequency, top exercises, PRs	period_days

Resources

Resource	Description
gym://exercise-catalog	Complete exercise catalog with muscle groups and equipment

Example prompts

• "How's my bench press progress looking?"
• "Am I training enough leg volume compared to upper body?"
• "Give me a summary of my last month"
• "Am I on track with my training plan?"
• "When was my last PR on squats?"
• "Suggest improvements to my push routine"

Setup

1. Install dependencies

cd gym-tracker-mcp
npm install

2. Configure environment

Copy .env.example to .env and fill in your credentials:

cp .env.example .env

SUPABASE_URL=https://your-project.supabase.co
SUPABASE_ANON_KEY=eyJ...
[email protected]
GYM_PASSWORD=your-password

Uses your existing Gym Tracker account credentials. All queries go through Supabase RLS — you can only access your own data.

3. Build

npm run build

4. Connect to your LLM client

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "gym-coach": {
      "type": "stdio",
      "command": "node",
      "args": ["/absolute/path/to/gym-tracker-mcp/dist/index.js"]
    }
  }
}

Claude Code

Add to .mcp.json in your project root or ~/.claude.json globally:

{
  "mcpServers": {
    "gym-coach": {
      "type": "stdio",
      "command": "node",
      "args": ["/absolute/path/to/gym-tracker-mcp/dist/index.js"]
    }
  }
}

Cursor

Add to .cursor/mcp.json in your project:

{
  "mcpServers": {
    "gym-coach": {
      "type": "stdio",
      "command": "node",
      "args": ["/absolute/path/to/gym-tracker-mcp/dist/index.js"]
    }
  }
}

Continue (VS Code / JetBrains)

Add to ~/.continue/config.json:

{
  "mcpServers": [
    {
      "name": "gym-coach",
      "command": "node",
      "args": ["/absolute/path/to/gym-tracker-mcp/dist/index.js"]
    }
  ]
}

Replace /absolute/path/to/ with the actual path to your gym-tracker-mcp directory.

Development

npm run dev    # Watch mode with tsx (hot reload)
npm run build  # Compile TypeScript to dist/
npm start      # Run compiled server

Architecture

gym-tracker-mcp/
├── src/
│   └── index.ts      # MCP server (~450 lines)
├── dist/              # Compiled JS (after npm run build)
├── .env               # Your credentials (not committed)
├── .env.example       # Template
├── package.json
└── tsconfig.json

• Transport: STDIO (maximum client compatibility)
• Auth: supabase.auth.signInWithPassword() with user's email/password
• Security: Uses anon key + RLS — each user can only access their own data
• Database: Reads directly from Gym Tracker's Supabase tables (workouts, routines, exercise_catalog, training_plans)

Tech stack

• Model Context Protocol SDK
• Supabase JS Client
• Zod (schema validation)
• TypeScript + Node.js