Boostcamp MCP Server

A Model Context Protocol (MCP) server for integrating with the Boostcamp fitness platform. It gives Claude Desktop and Claude Code seamless access to your training history, workout programs, custom exercises, and analytics.

Built on the boostcamp-api Python library — a wrapper around Boostcamp's private API.

Attribution: This project began as a fork of Alex-Keyes/boostcamp-mcp and builds on Alex-Keyes/boostcamp-api by Alex Keyes. It is now maintained as a standalone project. See Credits.

🚀 Quick Start

1. Installation

1. Clone this repository:

   git clone https://github.com/dcaslin/boostcamp-mcp.git
   cd boostcamp-mcp

2. Install dependencies with uv:

   uv sync

2. One-Time Authentication

Authentication runs through a standalone script so your credentials are never handled by the MCP client. From the project directory:

uv run login

Follow the prompts:

• Enter your Boostcamp email and password.

• The script authenticates against the Boostcamp API and saves your session token as BOOSTCAMP_AUTH_TOKEN in a local .env file (ignored by git).

Your email and password are only used to obtain the token and are never stored.

3. Register the Server with Claude

Claude Code (CLI):

claude mcp add boostcamp -- uv run --directory /path/to/your/boostcamp-mcp boostcamp-mcp

Claude Desktop: add this to your config file —

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "boostcamp": {
      "command": "uv",
      "args": [
        "run",
        "--directory",
        "/path/to/your/boostcamp-mcp",
        "boostcamp-mcp"
      ]
    }
  }
}

Important: Replace /path/to/your/boostcamp-mcp with the actual path to your clone. The server reads .env from that directory, so it must match where you ran uv run login. Restart Claude Desktop after editing the config.

4. Start Using

Once connected, ask Claude to use the tools directly, e.g.:

• "Show my Boostcamp profile" → get_my_profile

• "What programs am I enrolled in?" → list_enrolled_programs

• "Review my recent workouts" → get_training_history

• "What's my dashboard streak and totals?" → get_home_summary

Related MCP server: Arvo MCP Server

✨ Features

📊 Fitness Analytics

• Home Summary: total workouts, total weight moved, and current week streak.

• Volume Charts: training volume over time.

• Muscle Distribution: which muscle groups you've been targeting.

🏋️ Workout Management

• Program Details: full workout plans, including sets, reps, and coach notes.

• Enrolled Programs: track your progress in active training plans.

• Custom Exercises: access exercises you've manually created.

📚 Content & Discovery

• Program Catalog: search and list programs available on the platform.

• Blog Access: read the latest articles and training guides from the Boostcamp blog.

🛠️ Available Tools

timezone_offset is in minutes from UTC and defaults to -300.

Working with training history

get_training_history is built to keep responses small enough for any MCP client. By default it returns a summary of your 50 most recent workouts (date, program, exercises, and total volume) plus pagination metadata:

{
  "workouts": [ { "date": "2026-06-10", "program_name": "...",
                  "exercises": ["Squat (Barbell)", "..."],
                  "total_volume": 11570, "volume_unit": "lbs" } ],
  "pagination": {"page": 1, "page_size": 50, "total": 231,
                 "returned": 50, "has_more": true},
  "filters": {"start_date": null, "end_date": null, "detail": "summary"},
  "hint": "231 workouts match. Showing 50 (page 1, summary). Use page=2 ..."
}

• Filter by date: start_date / end_date as YYYY-MM-DD (inclusive), e.g. "my workouts since 2026-04-07".

• Page through history: bump page while pagination.has_more is true (page_size caps at 100 for summary, 25 for full).

• Get every set and rep: detail="full" adds per-exercise records with each set's weight, reps, target, and RPE. Use a small page_size here.

Supersets are flattened: each exercise inside a superset is listed individually (with its sets counted toward total_volume), and in full detail those exercises carry a superset id so the grouping is still visible.

🔧 Troubleshooting

Authentication Issues

If a tool returns an "Authentication Error" or your token has expired:

1. Re-run the login command from the project directory: uv run login

2. Restart your MCP client (Claude Desktop or Claude Code).

Session Management

• Sessions are stored in .boostcamp/session.pickle as {token, refresh_token}. The Firebase refresh token (not your password) is persisted, so an expired ID token is renewed automatically on the next request — no re-login needed for routine expiry.

• A BOOSTCAMP_AUTH_TOKEN is also saved to .env and used as a fallback for logins that predate refresh-token support (token-only, no auto-renewal).

• Upgrading: if you logged in before refresh-token support was added, run uv run login once to store the refresh token and enable auto-renewal.

• You only need to re-run uv run login if the refresh token itself becomes invalid (e.g. after a password change).

• Security Note: Never commit your .env or .boostcamp/ folder. They are included in .gitignore by default.

📄 License

Released under the MIT License.

🙏 Credits

• Original MCP server: Alex-Keyes/boostcamp-mcp by Alex Keyes.

• Underlying API library: Alex-Keyes/boostcamp-api by Alex Keyes, maintained here as dcaslin/boostcamp-api.