github-pr-review
Enables LLM clients to list, read, review, and comment on GitHub pull requests, turning an AI assistant into a fully capable code reviewer.
README
GitHub PR Review MCP Server
An MCP server that gives any LLM client the ability to list, read, review, and comment on GitHub pull requests — turning your AI assistant into a fully capable code reviewer.
Overview
This server wraps the GitHub REST API using FastMCP, exposing pull request operations as tools any MCP-compatible client can call. Connect it to Claude Desktop and you can ask Claude to review your open PRs, read diffs, post inline comments, approve, or request changes — all without leaving your conversation.
The server maintains a single authenticated httpx client across its lifetime, initialised on startup and shared across all tool calls via FastMCP's lifespan context.
Tools
| Tool | Description |
|---|---|
list_open_prs |
List all open PRs with author, title, draft status, and URL |
get_pr_diff |
Get file changes and unified diffs for a specific PR |
get_pr_comments |
Get all existing review comments on a PR |
post_review_comment |
Post an inline comment on a specific file and line |
approve_pr |
Approve a PR with an optional review message |
request_changes |
Request changes on a PR with a review message |
Note:
approve_prandrequest_changesrequire the authenticated user to be different from the PR author. GitHub does not permit self-review.
Project Structure
github-pr-mcp/
├── server.py # FastMCP instance, lifespan, tool registration
├── config.py # Environment variables and constants
├── tools/
│ ├── __init__.py
│ ├── prs.py # list_open_prs, get_pr_diff
│ └── reviews.py # get_pr_comments, post_review_comment, approve_pr, request_changes
├── .env # Your secrets (never committed)
├── .env.example # Template
├── requirements.txt
├── Dockerfile
└── README.md
Requirements
- Python 3.11+
- uv
- A GitHub Personal Access Token with
reposcope
Setup
1. Clone the repository:
git clone https://github.com/Festuskipkoech/github-pr-mcp.git
cd github-pr-mcp
2. Install dependencies:
uv sync
3. Configure your environment:
cp .env.example .env
Edit .env and fill in your values:
GITHUB_TOKEN=your_personal_access_token
GITHUB_OWNER=your_github_username
GITHUB_REPO=your_repository_name
Generate a token at GitHub Settings → Developer Settings → Personal Access Tokens. Enable full repo scope for read and write access.
Running Locally
uv run python server.py
Or with the FastMCP CLI on Streamable HTTP transport:
fastmcp run server.py:mcp --transport streamable-http
Testing with MCP Inspector
Start the server in one terminal:
fastmcp run server.py:mcp --transport streamable-http
Launch the Inspector in a second terminal:
npx -y @modelcontextprotocol/inspector
In the Inspector UI:
- Transport: Streamable HTTP
- URL:
http://127.0.0.1:8000/mcp - Click Connect
Recommended test sequence:
| Step | Tool | Input |
|---|---|---|
| 1 | list_open_prs |
(no arguments) |
| 2 | get_pr_diff |
{"pr_number": 1} |
| 3 | get_pr_comments |
{"pr_number": 1} |
| 4 | post_review_comment |
{"pr_number": 1, "body": "Looks good", "path": "file.py", "line": 1} |
Using with Claude Desktop
Add the following to ~/.config/Claude/claude_desktop_config.json:
{
"mcpServers": {
"github-pr-review": {
"command": "/home/your-user/.local/bin/uv",
"args": [
"run",
"--directory",
"/absolute/path/to/github-pr-mcp",
"python",
"server.py"
]
}
}
}
Use the full path to uv (find it with which uv). Credentials are read from the .env file automatically — no need to pass them in the config.
Restart Claude Desktop, then try:
List all open pull requests in my repo and show me the diff for PR number 1
Claude calls list_open_prs and get_pr_diff and returns structured results:

Full PR diff and file changes returned inline:

Running with Docker
docker build -t github-pr-mcp .
docker run --env-file .env -p 8000:8000 github-pr-mcp
For Streamable HTTP transport, update server.py to run with:
mcp.run(transport="streamable-http", host="0.0.0.0", port=8000)
Troubleshooting
| Error | Cause | Fix |
|---|---|---|
401 Unauthorized |
Invalid or expired token | Regenerate PAT on GitHub |
403 Forbidden |
Token lacks write scope | Enable full repo scope on your PAT |
404 Not Found |
Wrong owner or repo name | Check .env — use just the repo name, not the full URL |
422 Unprocessable on approve |
Self-review attempt | GitHub blocks approving your own PRs by design |
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。