A lightweight, sandboxed agent that uses an LLM to analyze, fix, and refactor Python code via tool calls. Built as a toy version of editor agents like Cursor/Zed/Claude Code—use cautiously.
- Bug detection and patch suggestions
- Optional refactors and style improvements
- Function-call tools for:
- Reading files
- Writing patches
- Running tests/commands (sandboxed)
- Configurable model/provider
- Treat this as a toy. Do not grant access to sensitive files.
- Run in a sandboxed environment (venv, containers).
- Commit before running the agent so you can revert.