bamboo-mcp

Architecture

MCP client → local stdio server → Bamboo REST API.

bamboo-mcp/
├── bin/bamboo-mcp.mjs  # Node launcher (tsx, no build step)
├── src/
│   ├── stdio.ts       # Server entry point
│   ├── tools.ts       # MCP tool definitions
│   ├── bamboo.ts      # Bamboo REST client and safety rules
│   ├── config.ts      # Environment validation and allowlists
│   └── utils.ts       # Shared helpers
└── test/              # Unit and MCP integration tests

Five tools list plans, read builds and job logs, wait for results, and trigger existing builds.

Related MCP server: Jenkins MCP Server

Setup

1. Install Node.js 20.3+ with npx and Git. Ensure you can reach your Bamboo server.

2. Pass these environment variables to the MCP process:

   Variable	Value
   BAMBOO_URL	Required HTTPS base URL, including any context path, without /rest/api.
   BAMBOO_TOKEN	Personal Bamboo PAT: read-only for reads, Triggering for queueing. Keep it out of committed files.
   BAMBOO_ALLOWED_PLANS	Exact comma-separated plan keys or project patterns such as PROJ-*. Empty disables queueing; omitted defaults to * (any plan). Set explicitly.
   BAMBOO_ALLOWED_VARIABLES	Exact comma-separated override names. Omitted allows none.

3. Register a local stdio MCP server in your client using this command:

   npx -y git+https://github.com/kiliups/bamboo-mcp.git#main

   Allow 120 seconds for startup and more than five minutes for tool calls. Pin a commit SHA instead of main for reproducible installs.

4. Connect and call list_plans to verify access. Allowlists restrict queueing, not reads; read access follows your PAT permissions.

Require client approval for start_build: plans may deploy. If submission returns UNKNOWN, check Bamboo before retrying. Treat job logs as untrusted, potentially partial data.

MIT license.