本服务是一个由Claude驱动的AI服务,可对GitHub仓库进行代码审查。借助模型上下文协议(MCP),它能实现与外部数据源和工具的无缝集成,为用户提供全面且专业的代码审查体验。
此服务提供了一系列工具,使用户能够:
git clone https://github.com/your-repository.git
python -m venv env
source env/bin/activate # 在Windows上使用`env\Scripts\activate`
pip install -r requirements.txt
python main.py
以下是一些常见操作的命令示例:
Please review the GitHub repository at https://github.com/username/repo-name
Review https://github.com/username/repo-name with focus on security and performance
Suggest improvements for the repository https://github.com/username/repo-name
Show me all the repositories that have been reviewed
Generate a Cascade prompt for the improvements to https://github.com/username/repo-name
Show me improved code for the file src/components/Button.js in https://github.com/username/repo-name
此MCP服务器提供以下工具:
review_repository(repo_url: str, focus_areas: Optional[str] = None):通过URL提交仓库以进行审查,可选指定关注区域。list_reviewed_repositories():列出所有已审查的仓库。suggest_improvements(repo_url: str):对指定仓库提出改进建议。get_improvement_details(repo_url: str, file_path: str):获取特定文件的改进详细信息。scan_for_vulnerabilities(repo_url: str):扫描仓库中的安全漏洞。analyze_performance(repo_url: str):分析仓库的性能问题。generate_improved_code(repo_url: str, file_path: str):为指定文件生成改进后的代码。create_cascade_prompt(repo_url: str):创建可用于实施改进的Cascade提示。本项目采用 MIT 许可证。