🤖 Pydantic DeepAgents: Claude Code-Style Agent Scaffold

Built on top of Pydantic AI, this project aims to help developers create Claude Code-style deep agents in Python. It integrates advanced features like tool-calling, sandboxed execution, multi-agent teams, skill management, and state checkpoints, while also supporting the Model Context Protocol (MCP). The interesting part is that it does not reinvent the wheel. By leveraging Pydantic AI, it focuses on the practical pain points of running complex agents, such as executing code safely within Docker sandboxes and managing long contexts and task states. This is not just a simple API wrapper; it is more like a tailored scaffold for deep research and coding agent scenarios. For developers looking to build autonomous agents with robust execution capabilities, it provides a well-structured reference implementation.