A Model Context Protocol server that exposes the GitLab REST v4 and GraphQL APIs as tools for AI assistants. It runs as a local binary over stdio (default) or HTTP.

This package is a thin launcher. The actual server is a prebuilt Go binary that ships inside a per-platform package; npm installs only the one matching your operating system and CPU, so there is nothing to compile and nothing is downloaded at install time.

Run without installing

npx @jmrp.io/gitlab-mcp-server

Most MCP clients are configured to launch the server this way. For example:

{
  "mcpServers": {
    "gitlab": {
      "command": "npx",
      "args": ["-y", "@jmrp.io/gitlab-mcp-server"],
      "env": {
        "GITLAB_URL": "https://gitlab.com",
        "GITLAB_TOKEN": "glpat-…"
      }
    }
  }
}

Install

npm install -g @jmrp.io/gitlab-mcp-server
gitlab-mcp-server --help

Configuration

GITLAB_TOKEN is the one required environment variable in stdio mode: a personal access token with the api scope (read_api is served the actions GitLab accepts from it). GITLAB_URL defaults to https://gitlab.com, so set it only for a self-managed instance. Every flag and variable (HTTP mode, OAuth, read-only and safe modes, tool surfaces, tiers) is documented in the configuration reference.

Updating

npm update -g @jmrp.io/gitlab-mcp-server. The server never replaces its own binary, on this or any other channel.

Supported platforms

Linux, macOS and Windows, on x64 and arm64. The Linux packages declare libc: ["glibc"]: the prebuilt binaries are PIE ELF executables that need the glibc dynamic loader, so npm skips them on musl distributions such as Alpine. Run the Docker image there, which is musl-based, or build from source. On any other platform the launcher exits with a message pointing to the release binaries and the option to build from source.