AI导哈

Python代码工具

一个用于在VSCode中运行Python脚本、进行代码检查、类型检查和安全性扫描的工具。

  • by zerolagtime
  • Python
  • 本地部署
  • 更新于
在 GitHub 查看 zerolagtime/mcp-server-python

使用说明

受信任的Python MCP工具

最小计算平台（MCP）工具 continue.dev VSCode中的插件，用于运行Python脚本、执行linting（使用 ruff)，类型检查（使用 ty 来自astral sh）和安全扫描（使用 bandit).它应该与其他工具配合得很好

特性

  • 在Docker容器中独立运行Python 3.11代码。
  • 使用以下命令检查Python代码样式和lint问题 ruff.
  • 使用检查类型错误 ty.
  • 使用执行安全分析 bandit.
  • 为每个会话使用隔离的临时目录进行缓存和文件管理。
  • 完全在Docker容器内运行，以确保沙盒并防止未经授权的文件系统访问。

快速入门

安装 uvx。不需要管理权限，但您需要将其永久添加到您的路径中（例如 ~/.local/bin)

按原样使用 uvx mcp-trusted-python.

如果代理始终可以使用额外的库，则将其放入需求文件中（例如 requirements-science.txt).然后，替换你喜欢的Python解释器，

uvx create --python python3.11 mcp-trusted-python
uvx shell mcp-trusted-python
uv pip install mcp-trusted-python
uv pip install -r requirements-science.txt

许可证

麻省理工学院

Python开发规则

将这些添加到工具的规则中以包含 *.py 文件夹。对于continue.dev，这些可以进入 ~/.continue/rules/trusted-python-rules.md.规则

## Test-Driven Development
- Write tests FIRST before implementation code
- Use `list_installed_packages()` to check available testing frameworks (pytest recommended)
- Run tests with `run_python()` to verify behavior before suggesting code
- Aim for >80% code coverage on business logic

## Code Quality & Validation
- Always `check_python()` and `security_scan()` before presenting code
- Use type hints for all function signatures (Python 3.11+)
- Keep cyclomatic complexity ≤3 per function (max 3 decision points/branches)
- Extract complex logic into well-named helper functions
- Write Pythonic, idiomatic code when it improves clarity

## SOLID Principles
- **Single Responsibility**: One function = one clear purpose
- **Open/Closed**: Use protocols/ABC for extensibility
- **Liskov Substitution**: Subtypes must be substitutable
- **Interface Segregation**: Small, focused interfaces over large ones
- **Dependency Inversion**: Depend on abstractions (protocols), inject dependencies

## Security & Configuration
- Use `python-dotenv` for sensitive values, never hardcode secrets
- Use stdlib `hashlib` and `secrets` for cryptography, avoid third-party crypto libs
- Validate all external inputs with type hints + runtime validation (pydantic recommended)
- Web applications always include CORS support and disable debugging at run time

## Architecture Patterns
- Separate concerns: handlers → services → repositories
- Use dataclasses/pydantic models for data structures
- Implement error handling at boundaries with custom exceptions
- Document public APIs with docstrings (Google/NumPy style)
- Prefer composition over inheritance

## Workflow
1. Query `list_installed_packages()` to verify available dependencies
2. Write failing test
3. Implement minimal code to pass test
4. Run `check_python()` and `security_scan()`
5. Refactor while keeping tests green
6. Present validated, tested code

相关 MCP