GitLab API MCP

What GitLab API MCP does

GitLab API MCP MCP server connects MCP-compatible agents to a GitLab instance through a Python-based API wrapper and an MCP interface. It covers common repository and project workflows, including branches, commits, projects, issues, merge requests, releases, tags, snippets, wikis, pipelines, jobs, environments, labels, milestones, groups, users, and members.

The server also exposes GitLab GraphQL operations, including schema discovery and raw queries or mutations. Additional tools support notes and comments, deploy tokens, protected branches, runners, packages, pipeline schedules, vulnerabilities, epics, and merge rules. Two ingestion tools can load GitLab projects and CI pipeline runs with jobs into epistemic-graph.

This belongs in the Version Control category for agents that need to read GitLab data or perform GitLab changes through natural-language workflows. The available operations still depend on the permissions of the GitLab credentials and the API endpoint being used.

How it works

GitLab API MCP MCP server offers two MCP tool layouts. The default condensed mode uses action-routed tools such as gitlab_projects, gitlab_issues, and gitlab_pipelines; each groups related operations behind a resource-specific entry point. verbose mode exposes individual tools for public API methods, such as creating a branch, accepting a merge request, or cancelling a project job. both enables both layouts.

Set MCP_TOOL_MODE to condensed, verbose, or both. Condensed mode is the default. In verbose mode, tools can be filtered with the verbose tag by using --tools tag:verbose or MCP_ENABLED_TAGS=verbose. Resource modules also have toggle variables, including PROJECTSTOOL, ISSUESTOOL, MERGE_REQUESTSTOOL, and PIPELINESTOOL.

The server includes api_request for arbitrary GitLab REST requests, along with GraphQL tools for raw queries, mutations, typed operations, and schema discovery.

Setup and configuration

Installation, deployment, API usage, CLI usage, and MCP usage are maintained in the project's official documentation. The package is published as gitlab-api, and the README identifies version 27.3.2.

MCP behavior can be configured through the shared ~/.config/agent-connector-sdk/config.json file or environment variables. Use MCP_TOOL_MODE to select the tool layout and MCP_ENABLED_TAGS=verbose to enable the documented verbose tag filter. Individual tool groups can be toggled with their corresponding environment variables from the MCP tool table.

Consult the official documentation before deployment, especially when connecting to a self-managed GitLab instance.

Tools and capabilities

The condensed tool surface includes grouped operations for:

• Projects, groups, namespaces, users, and members
• Branches, commits, tags, protected branches, and repositories
• Issues, epics, labels, milestones, notes, and merge requests
• Pipelines, pipeline schedules, jobs, environments, and runners
• Releases, packages, snippets, wikis, deploy tokens, and vulnerabilities
• REST requests, GraphQL queries, typed GraphQL operations, and schema discovery

The verbose surface maps individual public API methods to separate tools. Examples include creating projects, issues, branches, commits, environments, releases, pipeline schedules, and merge requests; approving or merging merge requests; managing project jobs; and adding comments.

Limitations and notes

It also does not guarantee that every GitLab API feature is represented beyond the listed tool surface. Review the generated live tool table and official documentation when selecting modules or planning permissions. The api_request and GraphQL capabilities can broaden access, but the material does not define their permission behavior separately from GitLab access controls.