Basecamp MCP Server

The first open-source MCP server for Basecamp 3 with multi-user and enterprise hosting support. Gives AI clients (Claude Desktop, Cursor, Codex, etc.) direct access to Basecamp via 148+ tools.

Why this project?

Every existing Basecamp MCP integration was single-user only — you had to run a local process per user with your own credentials. That doesn't work for teams or enterprise deployments.

This project solves that with a streamable-http mode: one server, one port, multiple users — each authenticating with their own Basecamp account via OAuth. Deploy it once for your entire organization.

Related MCP server: production-grade-mcp-agentic-system

Modes

Quick Start

Prerequisites

• Python 3.10+

• A Basecamp 3 account

• A Basecamp OAuth app — register at launchpad.37signals.com/integrations

Install

git clone https://github.com/Attri-Inc/basecamp-mcp-server.git
cd Basecamp-MCP-Server

# Using uv (recommended)
uv venv --python 3.12 venv && source venv/bin/activate
uv pip install -r requirements.txt && uv pip install mcp

# Or using pip
python setup.py

Configure

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

cp .env.example .env

Your Basecamp account ID is in the URL: https://3.basecamp.com/{account_id}/projects

Mode 1: stdio (Single-user)

Step 1 — Authenticate with Basecamp (one-time):

MCP_TRANSPORT=streamable-http ./venv/bin/python basecamp_fastmcp.py

Visit http://localhost:8001, click Connect Basecamp, complete OAuth. Tokens are saved to oauth_tokens.json. Stop the server.

Step 2 — Add to your MCP client config:

{
  "mcpServers": {
    "basecamp": {
      "command": "/full/path/to/venv/bin/python",
      "args": ["/full/path/to/basecamp_fastmcp.py"],
      "env": {
        "PYTHONPATH": "/full/path/to/project",
        "BASECAMP_ACCOUNT_ID": "your_account_id"
      }
    }
  }
}

Claude Desktop — auto-generate config:

python generate_claude_desktop_config.py

Then restart Claude Desktop completely.

Mode 2: streamable-http (Multi-tenant)

This is the core feature that makes this project different. A single server instance handles any number of users, each connected to their own Basecamp account. No per-user processes, no shared credentials.

How it works

User A ──► POST /mcp  (Bearer api_key_A) ──► validates key ──► uses User A's Basecamp tokens
User B ──► POST /mcp  (Bearer api_key_B) ──► validates key ──► uses User B's Basecamp tokens
User C ──► POST /mcp  (Bearer api_key_C) ──► validates key ──► uses User C's Basecamp tokens

Each user authenticates once via OAuth and receives a unique API key. From that point on, every MCP request they make is scoped to their own Basecamp account. Tokens are stored in PostgreSQL and auto-refreshed transparently.

Setup

Requires PostgreSQL. Add to .env:

MCP_TRANSPORT=streamable-http
DATABASE_URL=postgresql://user:pass@localhost:5432/basecamp_mcp
MCP_SERVER_URL=http://your-server:8001
MCP_HOST=0.0.0.0
MCP_PORT=8001

Start the server:

MCP_TRANSPORT=streamable-http ./venv/bin/python basecamp_fastmcp.py

User onboarding (per user, one-time)

1. User visits http://your-server:8001/

2. Clicks Connect Basecamp → redirected to 37signals OAuth

3. Authorizes the app → redirected back to your server

4. Server stores their Basecamp tokens in PostgreSQL and generates a unique API key

5. User copies their API key and adds it to their MCP client config:

{
  "mcpServers": {
    "basecamp": {
      "url": "http://your-server:8001/mcp",
      "headers": { "Authorization": "Bearer <their_api_key>" }
    }
  }
}

After this, the user's AI client has full access to their own Basecamp — and only their Basecamp. No other user's data is accessible.

Available Tools (148 total)

Architecture

Troubleshooting

Logs:

tail -f basecamp_fastmcp.log
tail -f ~/Library/Logs/Claude/mcp-server-basecamp.log  # Claude Desktop (macOS)

Contributing

See CONTRIBUTING.md.

License

MIT — see LICENSE.

Created and open-sourced by Attri