# GitLab for AI agents. Managed auth, ready-made tools.

> Review merge requests, file issues, read code and check pipelines in your users' GitLab projects.

- Page: https://www.agent200.com/tools/gitlab/
- Category: Developer tools
- Authentication: Managed OAuth
- Identifier in code: `gitlab`
- Create an API key: https://agent200.com/platform/apikeys/
- Updated: 2026-11-15

## Instructions for AI agents

Connector: GitLab, through Agent200. In code: "gitlab".
What it does: Review merge requests, file issues, read code and check pipelines in your users' GitLab projects.
Page: https://www.agent200.com/tools/gitlab/

1. Get an Agent200 API key. If the developer has none, send them to https://agent200.com/platform/apikeys/ and ask them to choose "Create an API key". Read the key from the AGENT200_API_KEY environment variable. Never ask for it in chat, and never write it into code, prompts or logs.
2. Sign-in is managed OAuth. Each end user approves access once, on GitLab's own page, through the authorization link Agent200 returns. Never ask a user for a GitLab password or token in chat.
3. Give every request a stable id for the end user. Agent200 uses only that user's own connection.
4. Read tools: gitlab_list_merge_requests, gitlab_get_merge_request, gitlab_search_issues, gitlab_get_file, gitlab_list_pipelines. Write tools: gitlab_create_issue, gitlab_create_note, gitlab_retry_pipeline. Name "gitlab" in the tools of a request; the model decides which tool to call.
5. Stay within the project's settings. If a tool is turned off or a call is refused, tell the user. Do not work around it.
6. Before a write, confirm with the user, unless they asked for that exact change.
7. For setup with a specific framework, follow its guide:
   https://www.agent200.com/tools/gitlab/framework/claude/
   https://www.agent200.com/tools/gitlab/framework/chatgpt/
   https://www.agent200.com/tools/gitlab/framework/claude-code/
   https://www.agent200.com/tools/gitlab/framework/muse/
   https://www.agent200.com/tools/gitlab/framework/grok-bot/
   https://www.agent200.com/tools/gitlab/framework/openai-dots/
   https://www.agent200.com/tools/gitlab/framework/claude-duet/
   https://www.agent200.com/tools/gitlab/framework/hermes-agent/
   https://www.agent200.com/tools/gitlab/framework/openclaw/

The code on the connector page is a conceptual example of the SDK. SDK reference: https://docs.igpt.ai/

## About

With the GitLab connector, your agent works in each user's own GitLab projects through Agent200: it reviews open merge requests, finds and files issues, reads code and checks how pipelines ran. Every user connects their own account through managed OAuth, and the agent acts with that user's project permissions, nothing more.

GitLab is a DevOps platform that brings source code, merge requests, issues and CI/CD pipelines into one application.

## GitLab tools

- **List merge requests** (`gitlab_list_merge_requests`, read): List a project's merge requests, filtered by state, author or reviewer.
- **Get a merge request** (`gitlab_get_merge_request`, read): Read a merge request's description, status and changes.
- **Search issues** (`gitlab_search_issues`, read): Find issues by text, label, assignee or state.
- **Create an issue** (`gitlab_create_issue`, write): Open an issue in a project, with labels and an assignee.
- **Add a comment** (`gitlab_create_note`, write): Comment on an issue or a merge request.
- **Read a file** (`gitlab_get_file`, read): Read a file from a repository at a branch, tag or commit.
- **List pipelines** (`gitlab_list_pipelines`, read): List a project's CI/CD pipelines and their status.
- **Retry a pipeline** (`gitlab_retry_pipeline`, write): Run a pipeline's failed jobs again.

## Limits you can set

**Turn single tools on or off.** Keep reading tools such as `gitlab_list_merge_requests` and switch off `gitlab_retry_pipeline` if your agent should never trigger CI.

**Restrict what the agent reaches.** Choose the endpoints and scopes a project may use. A read-only agent never needs write access, and a call outside them is refused.

## Get started

### Step 1: Set up the SDK

Add the Agent200 SDK to the codebase you already have and authenticate with your API key. The key covers every connector enabled for the project.

File `agent.js`:

```js
// One Agent200 API key covers every connector enabled for the project.
const agent200 = new Agent200({
  apiKey: process.env.AGENT200_API_KEY,
});
```

(Conceptual example of the Agent200 SDK.)

### Step 2: Let your user connect GitLab

Each end user authorizes their own GitLab account once. Agent200 runs the OAuth flow, handles the callback, stores the tokens and refreshes them.

File `connect.js`:

```js
// Each end user authorizes their own GitLab account, once.
const connection = await agent200
  .user("user_123")
  .connect("gitlab");

if (connection.authorizationUrl) {
  // Show this link to your user; Agent200 handles the rest.
  console.log(connection.authorizationUrl);
}
```

(Conceptual example of the Agent200 SDK.)

### Step 3: Run with the GitLab tools

Name the user, the model and the tools. The model decides which GitLab tools to call, and Agent200 executes them with that user's connection.

File `run.js`:

```js
const result = await agent200
  .user("user_123")
  .run({
    model: "openai/gpt-5.6-sol",
    input: "Which merge requests are waiting for my review",
    tools: ["gitlab"],
  });

console.log(result.output);
```

(Conceptual example of the Agent200 SDK.)

## Sign-in and permissions

**Who signs in.** Each of your users, once, on GitLab's own page. You never see or store their GitLab password.

**What access is requested.** GitLab lists the scopes on its consent screen, such as read-only API access or full API access when the agent writes. The agent reaches only the projects that user can already open.

## Example requests

- "Which merge requests are waiting for my review"
- "Why did the last pipeline on main fail"
- "Open an issue for the login timeout bug and assign it to me"
- "Summarize the changes in merge request 42"

## Guides: GitLab with your AI framework

- [GitLab + Claude](https://www.agent200.com/tools/gitlab/framework/claude.md): Connect GitLab to Claude through Agent200's MCP server: managed OAuth, ready-made tools and one API key.
- [GitLab + ChatGPT](https://www.agent200.com/tools/gitlab/framework/chatgpt.md): ChatGPT cannot send an API key to a custom MCP server, so this guide builds GitLab into your own agent with OpenAI's models and the Agent200 SDK.
- [GitLab + Claude Code](https://www.agent200.com/tools/gitlab/framework/claude-code.md): Have Claude Code build a GitLab agent with the Agent200 SDK: managed OAuth, ready-made tools and one API key.
- [GitLab + Muse](https://www.agent200.com/tools/gitlab/framework/muse.md): Connect GitLab to Muse through Agent200's MCP server: managed OAuth, ready-made tools and one API key.
- [GitLab + Grok Bot](https://www.agent200.com/tools/gitlab/framework/grok-bot.md): Connect GitLab to Grok Bot through Agent200's MCP server: managed OAuth, ready-made tools and one API key.
- [GitLab + OpenAI Dots](https://www.agent200.com/tools/gitlab/framework/openai-dots.md): Dots cannot send an API key to a custom MCP server, so this guide builds an always-on GitLab agent with OpenAI's models and the Agent200 SDK.
- [GitLab + Claude Duet](https://www.agent200.com/tools/gitlab/framework/claude-duet.md): Connect GitLab to Claude Duet through Agent200's MCP server: managed OAuth, ready-made tools and one API key.
- [GitLab + Hermes Agent](https://www.agent200.com/tools/gitlab/framework/hermes-agent.md): Connect GitLab to Hermes Agent through Agent200's MCP server: managed OAuth, ready-made tools and one API key.
- [GitLab + OpenClaw](https://www.agent200.com/tools/gitlab/framework/openclaw.md): Connect GitLab to OpenClaw through Agent200's MCP server: managed OAuth, ready-made tools and one API key.

## FAQ

### Who handles the GitLab OAuth flow and the tokens?

Agent200. It runs the authorization, handles the callback on its own domain, stores the tokens and refreshes them. You do not build or maintain an OAuth integration for GitLab.

### Does my agent act as the end user?

Within what that user authorized, yes. Every request names an end user, and Agent200 uses only that user's GitLab connection, never another user's. It cannot reach anything the user or GitLab has not authorized.

### Can I limit what my agent can do in GitLab?

Yes. You choose which services, endpoints and authorization scopes are allowed in your Agent200 configuration, and the agent works only within them.

## Related connectors

- [GitHub](https://www.agent200.com/tools/github.md): Read repositories and work with issues and pull requests as the connected user.
- [Bitbucket](https://www.agent200.com/tools/bitbucket.md): List and open pull requests, comment, read code and check pipelines in your users' Bitbucket workspaces.
- [Jira](https://www.agent200.com/tools/jira.md): Search, create and transition Jira issues for the connected user.
- [Linear](https://www.agent200.com/tools/linear.md): Create and update issues, and read projects and cycles in your users' Linear teams.

---

Agent200: https://www.agent200.com/ · API keys: https://agent200.com/platform/apikeys/ · Docs: https://docs.igpt.ai/
