An adapter that lets the Pi coding agent use MCP servers through one proxy tool, so idle tool definitions don't fill the context window.
zereight/gitlab-mcp
MCP server that gives AI assistants structured tools to work with GitLab projects, merge requests, issues, pipelines, wikis and more.
- Stars
- 2,089
- Forks
- 368
- Open issues
- 11
- Last commit
- yesterday
What is it?
gitlab-mcp is a Model Context Protocol server that connects AI clients such as Claude, Cursor, VS Code and Copilot to a GitLab instance. It exposes around 280 tools covering projects, merge requests, issues, pipelines, wikis, releases, tags and milestones, built on the GitLab REST and GraphQL APIs. By default only a lean core set of about 40 tools is loaded, and additional toolsets can be activated during a session to keep the model's tool list small.
The server runs over stdio for local clients, and over Streamable HTTP or SSE for shared or hosted deployments. Authentication options include a personal access token, a local OAuth2 browser flow, an MCP OAuth proxy for remote clients, and per-request token forwarding for multi-user setups. Permission modes (readonly, modify, full), tool allow and deny filters, and group or project restrictions limit what an agent can change.
For merge request review, a two-step flow lists changed files and then fetches diffs in batches. Optional JMESPath filtering and result compaction shrink large responses. The same binary also offers a command-line interface for full payloads such as diffs and job logs, and it works with GitLab.com as well as self-managed instances.
Key features
- 01Around 280 tools for projects, merge requests, issues, pipelines, wikis and releases
- 02Default lean toolset with on-demand activation of additional tools
- 03Stdio, SSE and Streamable HTTP transports
- 04Personal access token, OAuth2 and per-request remote authorization
- 05Readonly and modify permission modes with tool and project filters
- 06Two-step merge request review with batched diff retrieval
Use cases
- Asking an AI coding assistant to review a merge request's changed files and diffs
- Letting an agent check pipeline job logs or trigger CI actions in GitLab
- Sharing one hosted GitLab MCP endpoint across a team with per-user tokens
- Running a read-only GitLab integration for reporting or triage
Quick start
- 1
Install the binary from npm
npm install -g @zereight/mcp-gitlab - 2
Set your GitLab personal access token (read without echo)
read -rs GITLAB_PERSONAL_ACCESS_TOKEN && export GITLAB_PERSONAL_ACCESS_TOKEN - 3
Register the server with Claude Code
claude mcp add gitlab --transport stdio --scope local --env GITLAB_PERSONAL_ACCESS_TOKEN="$GITLAB_PERSONAL_ACCESS_TOKEN" --env GITLAB_API_URL=https://gitlab.com/api/v4 -- zereight-mcp-gitlab
Things to know
- Default toolset exposes only about 40 tools; others must be activated with discover_tools.
- The api toolset is risky because it bypasses allowlists and safety checks.
In collections
Topics
FAQ
- What is gitlab-mcp?
- MCP server that gives AI assistants structured tools to work with GitLab projects, merge requests, issues, pipelines, wikis and more. gitlab-mcp is a Model Context Protocol server that connects AI clients such as Claude, Cursor, VS Code and Copilot to a GitLab instance. It exposes around 280 tools covering projects, merge requests, issues, pipelines, wikis, releases, tags and milestones, built on the GitLab REST and GraphQL APIs. By default only a lean core set of about 40 tools is loaded, and additional toolsets can be activated during a session to keep the model's tool list small.
- Is gitlab-mcp free to use? What license does it have?
- gitlab-mcp is open source under the MIT license. Check the license terms before using it in commercial products.
- Is gitlab-mcp still maintained?
- Actively maintained. The last commit to gitlab-mcp was on Nov 6, 2026.
- What can I use gitlab-mcp for?
- Asking an AI coding assistant to review a merge request's changed files and diffs Letting an agent check pipeline job logs or trigger CI actions in GitLab Sharing one hosted GitLab MCP endpoint across a team with per-user tokens Running a read-only GitLab integration for reporting or triage
- How do I install gitlab-mcp?
- Install the binary from npm: npm install -g @zereight/mcp-gitlab Set your GitLab personal access token (read without echo): read -rs GITLAB_PERSONAL_ACCESS_TOKEN && export GITLAB_PERSONAL_ACCESS_TOKEN Register the server with Claude Code: claude mcp add gitlab --transport stdio --scope local --env GITLAB_PERSONAL_ACCESS_TOKEN="$GITLAB_PERSONAL_ACCESS_TOKEN" --env GITLAB_API_URL=https://gitlab.com/api/v4 -- zereight-mcp-gitlab
- Where can I get gitlab-mcp?
- The source code and installation instructions are on GitHub: https://github.com/zereight/gitlab-mcp
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Hosted MCP server giving AI agents YouTube transcripts, search, channel and playlist listings, and free new-upload tracking.
An MCP server that lets AI assistants like Claude and Cursor run web searches through a SearXNG instance you operate or trust.
A Python CLI and MCP server that lets coding agents and chats pull the right code context via reusable, rule-based file selection.
A step-by-step Go workshop that teaches you to build a coding agent on the Claude API, adding file, shell and search tools one at a time.
Unity Editor and runtime plugin and MCP server that lets AI agents like Claude, Cursor and Copilot operate Unity projects and run LLMs inside games.