Kepler MCP GitLab Server

Kepler MCP GitLab Server

Production-ready MCP server providing GitLab integration with OAuth authentication, enabling AI assistants to manage projects, issues, merge requests, branches, files, and commits across GitLab instances.

Category
访问服务器

README

Kepler MCP GitLab Server

A production-ready MCP (Model Context Protocol) server providing GitLab integration tools for AI assistants like Claude Desktop and ChatGPT.

How It Works

This MCP server acts as a bridge between AI assistants (ChatGPT, Claude, etc.) and your GitLab instance. Users authenticate via GitLab OAuth - just like connecting any third-party app.

┌─────────────────┐      ┌─────────────────────┐      ┌─────────────────┐
│   AI Assistant  │      │   MCP Server        │      │    GitLab       │
│   (ChatGPT,     │◄────►│   (Your Company     │◄────►│    (Your        │
│    Claude)      │      │    Hosted)          │      │    Instance)    │
└─────────────────┘      └─────────────────────┘      └─────────────────┘
        │                         │                          │
        │    1. User connects     │                          │
        │    ─────────────────►   │                          │
        │                         │   2. Redirect to GitLab  │
        │                         │   ─────────────────────► │
        │                         │                          │
        │                         │   3. User authorizes     │
        │                         │   ◄───────────────────── │
        │                         │                          │
        │    4. Connected!        │   5. API calls with      │
        │    ◄─────────────────   │      user's permissions  │
        │                         │   ◄────────────────────► │

For end users: Simply add the MCP server URL to their AI assistant and authorize with GitLab. No tokens or configuration needed.

For administrators: One-time setup to deploy the server and register the OAuth application with GitLab.

Features

  • GitLab Integration: Access projects, issues, merge requests, and more through MCP tools
  • OAuth 2.0 Authentication: Users authorize via GitLab - no tokens to manage
  • Per-User Permissions: Each user's GitLab permissions are respected
  • Multiple Transport Modes: stdio for local (Claude Desktop), SSE for HTTP (ChatGPT)
  • Secure Token Storage: Encrypted session storage using Fernet encryption
  • Docker Support: Production-ready container images

Available Tools

Once connected, users can ask their AI assistant to:

Tool Description
Projects
list_projects List accessible GitLab projects
get_project Get project details
search_projects Search for projects
get_project_languages Get language breakdown
Issues
list_issues List issues in a project
get_issue Get issue details
create_issue Create a new issue
update_issue Update an issue
close_issue / reopen_issue Change issue state
list_issue_comments List comments on an issue
add_issue_comment Add a comment
Merge Requests
list_merge_requests List MRs in a project
get_merge_request Get MR details
create_merge_request Create a new MR
update_merge_request Update an MR
merge_merge_request Merge an MR
approve_merge_request Approve an MR
get_merge_request_changes Get MR diff
list_merge_request_discussions List review threads
Repository
list_branches List repository branches
get_branch Get branch details
create_branch Create a new branch
delete_branch Delete a branch
list_tags List repository tags
get_tag Get tag details
create_tag Create a new tag
delete_tag Delete a tag
compare_branches Compare two branches/commits
list_repository_tree List files and directories
get_file Get file metadata
get_file_content Get file contents (decoded)
create_file Create a new file
update_file Update an existing file
delete_file Delete a file
get_file_blame Get file blame/history
list_commits List repository commits
get_commit Get commit details
get_commit_diff Get commit diff
cherry_pick_commit Cherry-pick a commit
get_commit_refs Get refs containing a commit
Utilities
get_current_user Get authenticated user info
get_gitlab_config Get server configuration

Deployment Guide (For Administrators)

This section is for IT administrators deploying the MCP server for their organization.

Prerequisites

  • Python 3.12+ or Docker
  • Access to your GitLab instance (admin or ability to create OAuth applications)
  • A server/host to run the MCP server (can be the same as GitLab or separate)

Understanding OAuth Application Credentials

When you deploy this server, you need to register it as an OAuth application with GitLab. This is similar to how Slack, GitHub, or any third-party integration works.

Why? GitLab needs to know:

  1. Who is asking (Client ID) - Identifies your MCP server
  2. That it's really you (Client Secret) - Proves the request is legitimate
  3. Where to send users back (Redirect URI) - After they authorize

These are application credentials, not user credentials. You set them up once, and then all users in your organization can connect without any additional configuration.

┌─────────────────────────────────────────────────────────────────────┐
│                    OAuth Flow Explained                              │
├─────────────────────────────────────────────────────────────────────┤
│                                                                      │
│  1. User clicks "Connect to GitLab" in ChatGPT                      │
│                          │                                           │
│                          ▼                                           │
│  2. MCP Server redirects to GitLab:                                 │
│     gitlab.com/oauth/authorize?client_id=ABC123&...                 │
│                          │                                           │
│                          │  ◄── GitLab checks: "Is ABC123 a valid   │
│                          │       registered application?"            │
│                          ▼                                           │
│  3. User sees GitLab login page, then:                              │
│     "Kepler MCP Server wants to access your account"                │
│     [Authorize] [Deny]                                              │
│                          │                                           │
│                          ▼                                           │
│  4. User clicks Authorize, GitLab redirects back with temp code     │
│                          │                                           │
│                          ▼                                           │
│  5. MCP Server exchanges code for tokens:                           │
│     POST gitlab.com/oauth/token                                     │
│     { client_id: ABC123, client_secret: XYZ789, code: ... }         │
│                          │                                           │
│                          │  ◄── GitLab verifies the secret matches  │
│                          │       the registered application          │
│                          ▼                                           │
│  6. GitLab returns access token, user is connected!                 │
│                                                                      │
└─────────────────────────────────────────────────────────────────────┘

Step 1: Register OAuth Application in GitLab

  1. Log into your GitLab instance as an admin (or any user for user-level apps)

  2. Navigate to:

    • Instance-wide (recommended): Admin Area → Applications
    • User-level: User Settings → Applications
  3. Click New Application and fill in:

    Field Value
    Name Kepler MCP GitLab Server
    Redirect URI https://your-mcp-server.com/oauth/callback
    Confidential ✅ Yes (checked)
    Scopes apiread_userread_repository
  4. Click Save application

  5. Copy and securely store:

    • Application ID (Client ID)
    • Secret (Client Secret) - shown only once!

Step 2: Configure the Server

Create a configuration file or set environment variables:

Option A: Environment Variables

# Required: GitLab Instance
export KEPLER_MCP_GITLAB_URL=https://gitlab.your-company.com

# Required: OAuth Application Credentials (from Step 1)
export KEPLER_MCP_OAUTH_USER_AUTH_ENABLED=true
export KEPLER_MCP_OAUTH_CLIENT_ID=your_application_id_here
export KEPLER_MCP_OAUTH_CLIENT_SECRET=your_secret_here
export KEPLER_MCP_OAUTH_REDIRECT_URI=https://your-mcp-server.com/oauth/callback
export KEPLER_MCP_OAUTH_SCOPE="api read_user read_repository"

# Required: OAuth Endpoints (adjust domain for your GitLab)
export KEPLER_MCP_OAUTH_AUTHORIZATION_URL=https://gitlab.your-company.com/oauth/authorize
export KEPLER_MCP_OAUTH_TOKEN_URL=https://gitlab.your-company.com/oauth/token
export KEPLER_MCP_OAUTH_USERINFO_URL=https://gitlab.your-company.com/api/v4/user

# Server Settings
export KEPLER_MCP_TRANSPORT_MODE=sse
export KEPLER_MCP_HOST=0.0.0.0
export KEPLER_MCP_PORT=8000

# Optional: Token Encryption (recommended for production)
# Generate key: python -c "from cryptography.fernet import Fernet; print(Fernet.generate_key().decode())"
export KEPLER_MCP_TOKEN_ENCRYPTION_KEY=your_generated_fernet_key
export KEPLER_MCP_TOKEN_STORE_PATH=/var/lib/kepler-mcp/tokens.json

Option B: .env File

Create a .env file in the project root with the same variables (without export).

GitLab.com vs Self-Hosted

Setting GitLab.com Self-Hosted
GITLAB_URL https://gitlab.com https://gitlab.your-company.com
OAUTH_AUTHORIZATION_URL https://gitlab.com/oauth/authorize https://gitlab.your-company.com/oauth/authorize
OAUTH_TOKEN_URL https://gitlab.com/oauth/token https://gitlab.your-company.com/oauth/token
OAUTH_USERINFO_URL https://gitlab.com/api/v4/user https://gitlab.your-company.com/api/v4/user

Step 3: Deploy the Server

Option A: Direct Python

# Create virtual environment
python3.12 -m venv .venv
source .venv/bin/activate

# Install
pip install -e .

# Run
python -m kepler_mcp_gitlab.cli serve --transport sse

Option B: Docker

# Build
docker build -t kepler-mcp-gitlab -f docker/Dockerfile .

# Run
docker run -d \
  --name kepler-mcp-gitlab \
  -p 8000:8000 \
  -e KEPLER_MCP_GITLAB_URL=https://gitlab.your-company.com \
  -e KEPLER_MCP_OAUTH_USER_AUTH_ENABLED=true \
  -e KEPLER_MCP_OAUTH_CLIENT_ID=your_client_id \
  -e KEPLER_MCP_OAUTH_CLIENT_SECRET=your_secret \
  -e KEPLER_MCP_OAUTH_REDIRECT_URI=https://your-mcp-server.com/oauth/callback \
  -e KEPLER_MCP_OAUTH_AUTHORIZATION_URL=https://gitlab.your-company.com/oauth/authorize \
  -e KEPLER_MCP_OAUTH_TOKEN_URL=https://gitlab.your-company.com/oauth/token \
  -e KEPLER_MCP_OAUTH_SCOPE="api read_user read_repository" \
  kepler-mcp-gitlab

Option C: Docker Compose

version: '3.8'
services:
  mcp-gitlab:
    build:
      context: .
      dockerfile: docker/Dockerfile
    ports:
      - "8000:8000"
    environment:
      KEPLER_MCP_GITLAB_URL: https://gitlab.your-company.com
      KEPLER_MCP_OAUTH_USER_AUTH_ENABLED: "true"
      KEPLER_MCP_OAUTH_CLIENT_ID: ${OAUTH_CLIENT_ID}
      KEPLER_MCP_OAUTH_CLIENT_SECRET: ${OAUTH_CLIENT_SECRET}
      KEPLER_MCP_OAUTH_REDIRECT_URI: https://your-mcp-server.com/oauth/callback
      KEPLER_MCP_OAUTH_AUTHORIZATION_URL: https://gitlab.your-company.com/oauth/authorize
      KEPLER_MCP_OAUTH_TOKEN_URL: https://gitlab.your-company.com/oauth/token
      KEPLER_MCP_OAUTH_SCOPE: "api read_user read_repository"
    restart: unless-stopped

Step 4: Verify Deployment

# Health check
curl https://your-mcp-server.com/health

# Expected response:
# {"status": "ok", "app_name": "Kepler MCP GitLab Server", "environment": "prod"}

# Test OAuth redirect (should redirect to GitLab)
curl -I https://your-mcp-server.com/oauth/authorize

# Expected: HTTP 302 with Location header pointing to GitLab

Step 5: Inform Your Users

Once deployed, share these instructions with your users:

Connecting to GitLab from ChatGPT

  1. In ChatGPT, go to Settings → Connections → Add MCP Server
  2. Enter the server URL: https://your-mcp-server.com/sse
  3. Click Connect - you'll be redirected to GitLab
  4. Log in to GitLab and click "Authorize"
  5. Done! You can now ask ChatGPT about your GitLab projects.

Try: "List my GitLab projects" or "Show open issues in project X"


Configuration Reference

All Environment Variables

Variable Description Default Required
Core
KEPLER_MCP_APP_NAME Application name Kepler MCP GitLab Server No
KEPLER_MCP_LOG_LEVEL Log level INFO No
KEPLER_MCP_ENVIRONMENT Environment (local/dev/stage/prod) local No
KEPLER_MCP_TRANSPORT_MODE Transport (stdio/sse) stdio No
KEPLER_MCP_HOST Server bind host 0.0.0.0 No
KEPLER_MCP_PORT Server bind port 8000 No
GitLab
KEPLER_MCP_GITLAB_URL GitLab instance URL https://gitlab.com Yes
OAuth (required for SSE mode)
KEPLER_MCP_OAUTH_USER_AUTH_ENABLED Enable OAuth false Yes*
KEPLER_MCP_OAUTH_CLIENT_ID OAuth Client ID - Yes*
KEPLER_MCP_OAUTH_CLIENT_SECRET OAuth Client Secret - Yes*
KEPLER_MCP_OAUTH_AUTHORIZATION_URL OAuth authorize endpoint - Yes*
KEPLER_MCP_OAUTH_TOKEN_URL OAuth token endpoint - Yes*
KEPLER_MCP_OAUTH_REDIRECT_URI OAuth callback URL - Yes*
KEPLER_MCP_OAUTH_SCOPE OAuth scopes - Yes*
KEPLER_MCP_OAUTH_USERINFO_URL User info endpoint - No
Token Storage
KEPLER_MCP_TOKEN_ENCRYPTION_KEY Fernet key for encryption - No**
KEPLER_MCP_TOKEN_STORE_PATH Path for token persistence - No**
Rate Limiting
KEPLER_MCP_RATE_LIMIT_REQUESTS_PER_MINUTE Max requests/min 60 No
KEPLER_MCP_RATE_LIMIT_BURST Burst size 10 No

* Required when OAUTH_USER_AUTH_ENABLED=true ** Required together if persistent sessions are needed


Development

Setup

# Clone and setup
git clone <repo>
cd kepler-mcp-gitlab-server

# Create virtual environment
python3.12 -m venv .venv
source .venv/bin/activate

# Install with dev dependencies
pip install -e ".[dev]"

Quality Checks

make lint        # Ruff linting
make typecheck   # Mypy type checking
make test        # Run tests
make coverage    # Tests with coverage report
make check-all   # All checks

Project Structure

src/kepler_mcp_gitlab/
├── cli.py              # Command-line interface
├── server.py           # FastMCP server setup
├── config.py           # Configuration management
├── context.py          # Request context and session management
├── application.py      # Tool registration
├── transport.py        # stdio/SSE transport handlers
├── security.py         # Auth strategies, token handling
├── gitlab/
│   ├── client.py       # GitLab API client
│   └── exceptions.py   # GitLab-specific errors
├── oauth/
│   ├── flows.py        # OAuth flow implementations
│   ├── pkce.py         # PKCE support
│   ├── session.py      # Session management
│   └── token_store.py  # Token persistence
└── tools/
    ├── projects.py     # Project tools
    ├── issues.py       # Issue tools
    ├── merge_requests.py # MR tools
    └── repository.py   # Repository tools (branches, tags, files, commits)

Security

  • OAuth 2.0 with PKCE: Industry-standard secure authentication
  • No stored passwords: Users authenticate directly with GitLab
  • Per-user permissions: Each user's GitLab access rights are respected
  • Encrypted tokens: Session tokens encrypted at rest with Fernet
  • Secret redaction: Sensitive values never appear in logs
  • Non-root containers: Docker images run as unprivileged user

Troubleshooting

OAuth Errors

"Invalid redirect URI"

  • Ensure the redirect URI in GitLab exactly matches your server's callback URL
  • Check for trailing slashes, http vs https

"Invalid client"

  • Verify Client ID is correct
  • Check if the OAuth application is still active in GitLab

"Invalid state"

  • State expired (default 10 min) - try the flow again
  • Server may have restarted during the OAuth flow

Connection Issues

Health check fails

# Check if server is running
curl http://localhost:8000/health

# Check logs
docker logs kepler-mcp-gitlab

GitLab API errors

# Test GitLab connectivity directly
curl -H "Authorization: Bearer <token>" https://gitlab.your-company.com/api/v4/user

License

MIT License - See LICENSE for details.

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选