健康数据MCP服务器

DataBody MCP Server

An MCP (Model Context Protocol) server that allows Claude to interact with your DataBody health tracking data using secure OAuth authentication.

Features

The MCP server provides 60+ tools for:

• Authentication - Secure OAuth2 login with PKCE support
• Health Summary - Get your daily macros, goals, body stats, and recent workouts
• Nutrition Logging - Log food items with full macro tracking
• Food Search - Search foods, barcodes, favorites, and recents
• Goal Management - Create and manage fitness goals with auto-calculation
• AI Coach - Chat with the AI nutrition coach
• Meal Suggestions - Get AI-powered meal suggestions based on remaining macros
• Photo Analysis - Analyze food photos to estimate macros
• Workout Tracking - Log and view workout history
• Lift Programs - Create and edit ordered workout days, moves and set targets
• Set Tracking - Log actual reps and weight, manage sessions, notes and variations, and review performance and progression
• Notes & Memory - Save preferences and dietary restrictions
• Households - Share data with family for coordinated meal planning
• Chat Threads - Manage conversation history with the AI coach

Installation

Choose one of these installation methods:

Option A: Download MCPB Bundle (Recommended)

The easiest option. Download a single file that contains everything pre-bundled. No npm or Node.js required.

1. Download the latest .mcpb file from GitHub Releases
2. Double-click the file or drag it onto Claude Desktop
3. Done! The server is installed and configured automatically.

Option B: Install via npm

If you have Node.js 20+ installed:

npm install -g @databody/mcp

Option C: Build from Source

For developers who want to customize or contribute:

# Clone the repository
git clone https://github.com/databody-ai/databody-mcp.git
cd databody-mcp

# Install dependencies and build
npm install
npm run build

Configuration

If you installed via MCPB

No manual configuration needed! The server is configured automatically when you install the .mcpb file.

If you installed via npm or source

Add to your Claude Desktop config file:

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

For npm install:

{
  "mcpServers": {
    "databody": {
      "command": "npx",
      "args": ["-y", "@databody/mcp"],
      "env": {
        "DATABODY_API_URL": "https://databody.ai"
      }
    }
  }
}

For source install:

{
  "mcpServers": {
    "databody": {
      "command": "node",
      "args": ["/path/to/databody-mcp/dist/index.js"],
      "env": {
        "DATABODY_API_URL": "https://databody.ai"
      }
    }
  }
}

Replace /path/to/databody-mcp with the actual path where you cloned the repository.

Authentication

When you first use a DataBody tool, you'll need to authenticate. Use the authenticate tool:

Browser-based OAuth (recommended):

Use the authenticate tool with method "browser"

This will:

1. Start a local callback server on port 8787
2. Display a URL to open in your browser
3. After you log in and authorize, the token is saved automatically

Password-based (alternative):

Use the authenticate tool with method "password", email "your@email.com", and password "yourpassword"

Tokens are stored securely in ~/.databody_token.json with restricted permissions.

Environment Variables

The MCP server shares its OAuth client and default token file with the DataBody CLI. A CLI sign-in can be reused by MCP on the same machine. Use the same API URL and token-file override for both clients, and keep credential files private.

Available Tools

Authentication

• authenticate - Login via browser OAuth or password
• logout - Clear stored tokens
• auth_status - Check login status

User Profile

• get_user_profile - Get profile settings
• update_user_profile - Update name, height, activity level, etc.
• change_password / change_email - Account management

Health & Summary

• get_health_summary - Dashboard with macros, goals, stats, workouts
• get_health_history - Historical trends for weight, body fat, etc.

Nutrition

• get_today_nutrition - Today's food logs
• get_nutrition_history - Date range history
• log_food - Log meals with macro tracking
• update_nutrition_log / delete_nutrition_log - Manage logs
• add_nutrition_item / delete_nutrition_item - Manage items

Food Search

• search_food - Search FatSecret, USDA, OpenFoodFacts
• get_food_details - Detailed nutrition info
• barcode_lookup - Scan product barcodes
• get_favorites / add_favorite / remove_favorite - Manage favorites
• get_recents - Recently logged foods

AI Features

• ai_chat - Chat with the nutrition coach
• get_meal_suggestions - AI meal recommendations
• analyze_photo - Identify foods from photos
• parse_food_description - Parse text to structured data
• expand_meal_ingredients - Break meals into ingredients
• get_daily_greeting - Personalized daily greeting
• get_token_usage - AI usage statistics

Chat Threads

• list_chat_threads / get_chat_thread - View conversations
• create_chat_thread / delete_chat_thread - Manage threads
• generate_thread_title - AI-generate thread titles
• get_chat_history - Get messages for a thread

Goals

• get_current_goal / list_goals / get_goal - View goals
• create_goal / update_goal / delete_goal - Manage goals
• calculate_goal - Recalculate based on current stats

Workouts

• get_recent_workouts / list_workouts / get_workout - View workouts
• create_workout / update_workout / delete_workout - Manage workouts

Notes & Memory

• get_notes - View saved preferences
• create_note / update_note / delete_note - Manage notes

Households

• get_households / get_household_summary - View households
• create_household / update_household / delete_household - Manage
• list_household_members / remove_household_member - Manage members
• suggest_meal_to_member - Create pending logs for family members

Household Invites

• list_household_invites / list_pending_invites - View invites
• create_household_invite / cancel_household_invite - Send invites
• accept_invite / decline_invite / accept_invite_by_token - Respond

Structured Workouts

The 37 structured workout tools cover reusable moves, complete programs, sessions, individual sets, notes, variations, performance history and progression. Existing activity-summary and saved-meal tools remain available. The complete nested input schemas live in src/workout-tools.json and ship in dist/workout-tools.json. The repository builds independently of the Rails checkout.

Example Usage

DataBody Lift programs

The server exposes 37 structured workout tools from the shared API/Coach contract. create_workout_program creates the complete program in one call; creating library moves separately is optional. Read list_workout_programs before creating one to avoid accidental duplicates. Use get_workout_program before editing and supply the returned expected_revision. Cancel an empty session with cancel_workout_session; remove a completed session with delete_workout_session. Use return_to_workout_rest to recover from an accidental Next, or replace_workout_session_move with reviewed targets to substitute only the remaining sets of a move in this workout. set_workout_session_notes saves notes on an active or completed session; set_workout_move_notes saves reusable setup cues using a session move occurrence. Both require the current session revision and accept up to 2,000 characters, with an empty string to clear. update_workout_move also accepts notes. Read get_workout_move_performance for exercise charts/notes and paginated actual workouts; read get_workout_session_performance for last-time sets, historical records and matching-workout comparisons. These two performance tools are read-only.

Example create_workout_program arguments (illustrative targets; replace them with the user's supplied plan):

{
  "name": "Strength",
  "days": [{
    "id": "upper-a", "name": "Upper A",
    "moves": [{
      "id": "bench-press", "name": "Bench Press", "unit": "lb",
      "load_basis": "total", "increment": 5,
      "sets": [
        {"reps_min": 6, "reps_max": 8, "weight": 135, "rest_seconds": 90},
        {"reps_min": 6, "reps_max": 8, "weight": 135, "rest_seconds": 90},
        {"reps_min": 6, "reps_max": 8, "weight": 135, "rest_seconds": 90}
      ]
    }]
  }]
}

Array order determines the exercise and set sequence. A numeric move_id reuses a library move; a string occurrence id identifies its placement. Keep occurrence IDs when editing; use new IDs when duplicating. Supplying days replaces the entire day list, so retain all days/moves that should remain. Changes apply to future workouts; use session tools to change a workout already in progress.

For set logging, use start_workout_session, read its queue and set IDs, then log_workout_set with actual reps/weight and the current revision. Reuse a caller-generated client_id when retrying session starts, imports or commands. Never log planned reps as completed facts. get_workout_progression recommends increases; apply_workout_progression applies only selected changes.

Use remove_workout_session_set or remove_workout_session_move to remove pending work while preserving completed actuals. Set variation_ids through update_workout_move to link reusable alternatives; an empty array clears the links.

Restart the MCP connection after updating to refresh tool discovery.

Other examples

Once configured and authenticated, you can ask Claude things like:

• "What have I eaten today?"
• "Log a lunch: grilled salmon (350 cal, 40g protein, 0g carbs, 20g fat)"
• "What are my remaining macros for today?"
• "Search for nutrition info on greek yogurt"
• "What meal suggestions do you have for dinner?"
• "Show me my recent workouts"
• "What's my current fitness goal?"
• "Remember that I'm lactose intolerant"
• "Plan dinner for my family tonight"

Token Management

• Access tokens expire after 1 hour
• Refresh tokens are used automatically to get new access tokens
• Tokens are stored in ~/.databody_token.json with 600 permissions
• Use the logout tool to clear stored tokens
• Use the auth_status tool to check if you're logged in

Development

Run in development mode with hot reload:

npm run dev

Build for production:

npm run build
npm start

Security Notes

• Uses OAuth 2.0 with PKCE for secure authentication
• No secrets stored in the MCP configuration
• Tokens are stored with restricted file permissions
• Automatic token refresh when expired
• Browser-based flow is recommended over password flow

Troubleshooting

"Not authenticated": Run the authenticate tool to login.

"Session expired": Your token expired and couldn't be refreshed. Re-authenticate.

OAuth callback fails: Make sure port 8787 is available, or set DATABODY_CALLBACK_PORT to a different port.

Connection refused: Ensure the DataBody API is reachable at the configured URL.

Links

• DataBody App
• MCP Documentation
• API Documentation
• GitHub Issues