Coordinator
MCP-Coordinator transforms MCP servers into importable Python libraries, implementing Anthropic's progressive disclosure pattern to enable local data chaining and reduce token usage in AI agent workflows. The framework automatically discovers server capabilities, generates typed Python wrappers for all tools, and provides secure code execution through smolagents with multiple backend support including Docker, E2B, and local execution. Built with a hybrid skills system that combines database storage for dynamic agent-created skills with file-based fallback for version-controlled code, it enables AI agents to accumulate reusable capabilities while maintaining developer workflow compatibility.
