GitLab MCP Server
Enables LLMs to interact with GitLab via MCP, providing access to issues, merge requests, pipelines, and repository files for browsing, searching, and reading details.
README
GitLab MCP Server
An MCP (Model Context Protocol) server that allows LLMs to interact with GitLab. Access issues, merge requests, pipelines, and repository files directly from Claude Code or other MCP-compatible clients.
Features
- Browse and search projects
- View issues and merge requests (including your own and team-labeled)
- Read MR diffs and comments
- Check pipeline status and job logs
- Browse repository files
Setup
1. Install dependencies
make install
# or
uv sync
2. Create GitLab Personal Access Token
- Go to GitLab → Avatar (top-right) → Edit profile
- Left sidebar → Access Tokens
- Click Add new token
- Configure:
- Token name: e.g., "MCP Server"
- Expiration date: choose as needed
- Scopes:
read_api(required)read_repository(for file content access)
- Click Create personal access token
- Copy the token (won't be shown again)
Direct URL: https://gitlab.com/-/user_settings/personal_access_tokens
3. Configure environment
cp .env.example .env
Edit .env with your values:
GITLAB_URL=https://gitlab.com
GITLAB_TOKEN=your-private-token-here
DEFAULT_ASSIGNEE=me
Configuration
| Variable | Description | Default |
|---|---|---|
GITLAB_URL |
Your GitLab instance URL | https://gitlab.com |
GITLAB_TOKEN |
Personal access token | (required) |
DEFAULT_ASSIGNEE |
Username for "my_*" tools filters | me |
Usage
Run directly
uv run python main.py
Run in background
make start # Start server
make stop # Stop server
make restart # Restart server
make status # Check if running
make logs # Tail server logs
Add to Claude Code
Create .mcp.json in your project root:
{
"mcpServers": {
"gitlab": {
"command": "uv",
"args": ["run", "python", "main.py"],
"cwd": "/path/to/gitlab-mcp-server"
}
}
}
Or use the CLI:
claude mcp add gitlab -- uv run python main.py
Verify with /mcp command inside Claude Code.
Test the server
# Test with MCP Inspector
npx @modelcontextprotocol/inspector uv run python main.py
# Test API connection
uv run python -c "
from tools.projects import get_client
with get_client() as client:
r = client.get('/user')
print('Authenticated as:', r.json().get('username'))
"
Available Tools
Projects
| Tool | Description |
|---|---|
list_projects |
List accessible projects (search, filter by ownership/membership) |
get_project |
Get project details by path |
Issues
| Tool | Description |
|---|---|
list_issues |
List issues in a project (filter by state, labels, search) |
my_issues |
List issues assigned to you (uses DEFAULT_ASSIGNEE) |
team_issues |
List issues with a team label (default: "RE") |
get_issue |
Get issue details including description |
get_issue_comments |
Get comments/notes on an issue |
Merge Requests
| Tool | Description |
|---|---|
list_merge_requests |
List MRs in a project (filter by state, branches, search) |
my_merge_requests |
List MRs assigned to you |
my_merge_requests_to_review |
List MRs where you are a reviewer |
team_merge_requests |
List MRs with a team label (default: "RE") |
get_merge_request |
Get MR details including description |
get_merge_request_changes |
Get MR file diffs |
get_merge_request_comments |
Get MR comments/discussions |
Pipelines
| Tool | Description |
|---|---|
list_pipelines |
List pipelines in a project (filter by status, branch) |
get_pipeline_jobs |
Get jobs for a pipeline |
get_job_log |
Get job log output |
Repository
| Tool | Description |
|---|---|
list_repository_tree |
List files/directories in a repo |
get_file_content |
Get file content from a repo |
Project Structure
gitlab-mcp-server/
├── main.py # Entry point
├── client.py # HTTP client & settings
├── models.py # Pydantic models
├── tools/
│ ├── projects.py # Project tools
│ ├── issues.py # Issue tools
│ ├── merge_requests.py # MR tools
│ ├── pipelines.py # Pipeline tools
│ └── repository.py # Repository tools
├── .env.example # Environment template
└── Makefile # Run commands
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。