Feedback issue: GitLab MCP Server

Welcome to the GitLab MCP Server General Availability feedback issue

Feature maturity status

Beta, with general availability planned for GitLab 19.5

The purpose of this feedback issue is to collect your experiences using the GitLab MCP server: connecting it, finding the right tool, completing real work with it, and governing it. Our goal is to understand what's working, what's broken, and what's missing as we move through general availability and decide what to build next.

This issue supersedes #561564 (closed), which covered the experimental and beta periods and is now closed.

What is the GitLab MCP server?

The GitLab MCP server exposes GitLab as a set of tools that any MCP-compatible agent or IDE can call. Your agent reads merge requests, searches code, commits changes, drives pipelines, and triages vulnerabilities without leaving the tool you already work in.

This is the external agent → GitLab direction of MCP: the agent is the client, and GitLab is the context source and action provider. For the opposite direction, GitLab agents reaching out to external services, see #593219.

Common clients include Claude Code, Claude Desktop, Cursor, GitHub Copilot in VS Code, Gemini Code Assist and Gemini CLI, Kiro, OpenAI Codex, and Zed.

What's in scope for general availability?
  • 52 tools, 39 of them added since 19.0, covering merge requests, work items, commits and branches, repository browsing, pipelines and job logs, search, vulnerabilities, and Duo Agent Platform sessions
  • Toolsets, so you choose which groups of tools load. core, merge_requests, work_items, repository and ci are on by default; duo_agent_platform, wikis and code_security are opt-in (docs)
  • Tool governance, set at the top-level group and optionally tightened per project (docs)
  • OAuth authentication and HTTP transport

Known gaps we are still working on:

  • Replying to or resolving a specific merge request discussion thread
  • Searching GitLab documentation
  • Release and tag writes, and reading CI artifacts
  • Invoking Duo Agent Platform agents and flows over MCP

🎯 Feedback we're especially interested in

If you have tried our updated MCP Server and still currently use glab, the zereight community MCP server, or direct API calls

Many teams reach for a CLI, a community MCP server, or raw REST and GraphQL instead of ours. We want to understand that choice!

  1. What does your current approach do better? Coverage, speed, output shape, reliability, something else?
  2. What would have to be true for the GitLab MCP server to replace it?
  3. Did you try ours and go back? Tell us where it fell down. This is the most useful feedback we can get.

For developers, tool coverage and tool selection

  1. Did the agent find the right tool? If it picked the wrong one, called something that doesn't exist, or gave up and used a shell command, tell us the tool name it reached for.
  2. Which workflow could you not finish? Name the step where you had to leave MCP and do something by hand.
  3. What's missing entirely? Which GitLab operation do you need that has no tool?
  4. Output quality: is a tool returning too much, too little, or the wrong shape for your agent's context window?

For admins, governance and access

  1. Governance: do top-level group rules with per-project overrides fit how you need to control agent access, or do you need something different?
  2. Access: was enabling the MCP server for your group or instance clear? Where did you get stuck?
  3. Limits: what rate limit behavior do you need, and what would you want to see when a limit is hit?
  4. Audit and visibility: what do you need to see about which agents called which tools, and when?

For GitLab Self-Managed and Dedicated

  1. Was it clear whether the MCP server is available on your deployment? If the docs led you wrong, say so and point at the page.
  2. What differs from GitLab.com in a way that blocked you?

📝 How to give feedback

  1. Check existing threads: review comments below to see if your issue is already reported. Add a 👍 or comment to show support.
  2. Start a new thread: use a descriptive title so others can quickly understand the focus of your thread.
  3. Include context: tell us your GitLab version and whether you're on GitLab.com, Self-Managed, or Dedicated; which MCP client you were using; which tool was involved; what you were trying to do; and what happened. Error messages, the tool name your agent called, and screenshots are especially helpful.

🤝 What you can expect from us

  1. We will read all feedback while this issue remains open
  2. We will prioritize fixes and iterations based on feedback patterns
  3. We will create issues for reproducible bugs and link them back to your thread
  4. We may reach out to invite you into design partnership conversations as we plan the next iteration

👏 Thank you for helping us build a better agent platform. Your feedback directly shapes what we build next.

Edited by Amanda Rueda