GitLab MCP Server
A Model Context Protocol (MCP) server for GitLab integration, written in TypeScript. This server allows you to interact with GitLab merge requests, commits, pipelines, and jobs through MCP tools.
README
GitLab MCP Server
A Model Context Protocol (MCP) server for GitLab integration, written in TypeScript. This server allows you to interact with GitLab merge requests, commits, pipelines, and jobs through MCP tools.
Features
- 🔍 Get merge request details (title, description, metadata)
- 📝 Create merge requests in GitLab projects
- 📄 Retrieve merge request diffs
- 📋 List merge requests for a project
- 🔀 Get pipelines for a merge request
- 🧰 Get jobs for a pipeline
- 🚨 Get failed jobs by pipeline or merge request
- 📜 Retrieve job logs (trace)
- 📦 Retrieve job artifacts metadata and download links
- 🔐 Supports self-hosted GitLab instances
- 🔑 Token-based authentication via environment variables
Installation
- Clone or download this repository
- Install dependencies:
npm install - Build the project:
npm run build
Configuration
Set the following environment variables:
Required
GITLAB_HOST: Your GitLab instance URL (e.g.,https://gitlab.example.comorhttps://gitlab.com)GITLAB_API_TOKEN: Your GitLab access token
Optional (for multiple instances)
MCP_INSTANCE_ID: Unique identifier for the server instance (useful for logging and identification)
You can copy .env.example to .env and fill in your values:
cp .env.example .env
# Edit .env with your GitLab configuration
Getting a GitLab Access Token
- Go to your GitLab instance
- Navigate to User Settings → Access Tokens
- Create a new token with
apiscope - Copy the token and set it as
GITLAB_API_TOKEN
Usage
Running the Server
Single Instance
# Development mode
npm run dev
# Production mode
npm run build && npm start
Multiple Instances
Each instance runs as a separate process with its own configuration:
# Build first
npm run build
# Instance 1 (in terminal 1)
GITLAB_HOST=https://gitlab.example.com GITLAB_API_TOKEN=token1 MCP_INSTANCE_ID=instance1 npm start
# Instance 2 (in terminal 2)
GITLAB_HOST=https://gitlab.com GITLAB_API_TOKEN=token2 MCP_INSTANCE_ID=instance2 npm start
# Instance 3 (in terminal 3)
GITLAB_HOST=https://gitlab.internal.com GITLAB_API_TOKEN=token3 MCP_INSTANCE_ID=instance3 npm start
Each instance can connect to different GitLab hosts with different tokens, allowing multiple tools to work with different GitLab environments simultaneously.
Available Tools
get_merge_request
Get detailed information about a specific merge request, optionally including the diff/changes.
Parameters:
projectId(string): GitLab project ID or path (e.g., "group/project" or "123")mergeRequestIid(number): Merge request internal ID (IID)includeDiff(optional boolean): Whether to include the diff/changes in the response (default: false)
The response includes automation-friendly metadata such as id, iid, draft, source_project_id, target_project_id, labels, assignees, reviewers, diff_refs, and sha. When includeDiff is enabled, the response also includes diff_overflow.
create_merge_request
Create a merge request in a GitLab project.
Parameters:
projectId(string): GitLab project ID or pathsourceBranch(string): Source branch nametargetBranch(string): Target branch nametitle(string): Merge request titledescription(optional string): Merge request descriptionlabels(optional string[]): Labels to applyassigneeId/assigneeIds(optional number / number[]): Assignee user IDsreviewerId/reviewerIds(optional number / number[]): Reviewer user IDsremoveSourceBranch(optional boolean): Remove the source branch on mergeallowCollaboration(optional boolean): Allow upstream members to pushallowMaintainerToPush(optional boolean): Allow maintainers to pushsquash(optional boolean): Suggest squashing commits on mergetargetProjectId(optional number): Target project ID for cross-project MRsdraft(optional boolean): Prefix the title withDraft:
list_merge_requests
List merge requests for a project.
Parameters:
projectId(string): GitLab project ID or pathstate(optional string): Filter by state ("opened", "closed", "merged", "all"). Default: "opened"
get_commit_diff
Get commit details and diff using a full GitLab commit URL.
Parameters:
commitUrl(string): Full GitLab commit URL
get_merge_request_pipelines
Get pipelines for a merge request.
Parameters:
projectId(string): GitLab project ID or pathmergeRequestIid(number): Merge request internal ID (IID)
get_pipeline_jobs
Get jobs for a pipeline.
Parameters:
projectId(string): GitLab project ID or pathpipelineId(number): Pipeline ID
get_failed_jobs
Get failed jobs either directly by pipeline ID or from the latest pipeline of a merge request.
Parameters:
projectId(string): GitLab project ID or pathpipelineId(optional number): Pipeline IDmergeRequestIid(optional number): Merge request internal ID (IID)
Provide either pipelineId or mergeRequestIid.
get_job_log
Get job log (trace) for a job.
Parameters:
projectId(string): GitLab project ID or pathjobId(number): Job ID
get_job_artifacts
Get job artifacts metadata and download links for a job.
Parameters:
projectId(string): GitLab project ID or pathjobId(number): Job ID
Example Usage with MCP Client
// Get merge request details only
await callTool("get_merge_request", {
projectId: "mygroup/myproject",
mergeRequestIid: 42
});
// Get merge request details with diff
await callTool("get_merge_request", {
projectId: "123",
mergeRequestIid: 42,
includeDiff: true
});
// Create a merge request
await callTool("create_merge_request", {
projectId: "mygroup/myproject",
sourceBranch: "feature/new-e2e-test",
targetBranch: "main",
title: "Add login dashboard E2E coverage",
description: "## Summary\n- add a generated Playwright test\n- validate the scenario locally",
labels: ["qa", "e2e"],
removeSourceBranch: true
});
// List open merge requests
await callTool("list_merge_requests", {
projectId: "mygroup/myproject",
state: "opened"
});
// List pipelines for an MR
await callTool("get_merge_request_pipelines", {
projectId: "mygroup/myproject",
mergeRequestIid: 42
});
// Get jobs from pipeline
await callTool("get_pipeline_jobs", {
projectId: "mygroup/myproject",
pipelineId: 123456
});
// Get failed jobs from latest MR pipeline
await callTool("get_failed_jobs", {
projectId: "mygroup/myproject",
mergeRequestIid: 42
});
// Get raw CI job log
await callTool("get_job_log", {
projectId: "mygroup/myproject",
jobId: 987654
});
// Get artifact metadata and links
await callTool("get_job_artifacts", {
projectId: "mygroup/myproject",
jobId: 987654
});
Multiple Instance Setup
When running multiple instances, each tool/client can connect to a different instance:
MCP Client Configuration Example
{
"mcpServers": {
"gitlab-company": {
"command": "node",
"args": ["path/to/gitlab-mcp/dist/index.js"],
"env": {
"GITLAB_HOST": "https://gitlab.example.com",
"GITLAB_API_TOKEN": "company_token",
"MCP_INSTANCE_ID": "company"
}
},
"gitlab-personal": {
"command": "node",
"args": ["path/to/gitlab-mcp/dist/index.js"],
"env": {
"GITLAB_HOST": "https://gitlab.com",
"GITLAB_API_TOKEN": "personal_token",
"MCP_INSTANCE_ID": "personal"
}
},
"gitlab-internal": {
"command": "node",
"args": ["path/to/gitlab-mcp/dist/index.js"],
"env": {
"GITLAB_HOST": "https://gitlab.internal.company.com",
"GITLAB_API_TOKEN": "internal_token",
"MCP_INSTANCE_ID": "internal"
}
}
}
}
This allows different tools to connect to different GitLab instances simultaneously, each with their own authentication and configuration.
Development
The project uses TypeScript and the MCP SDK. Key files:
src/index.ts: Main MCP server implementationsrc/gitlab-client.ts: GitLab API client wrapperpackage.json: Dependencies and scriptstsconfig.json: TypeScript configuration
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。