Mcp Server Code Execution Mode

The Mcp Server Code Execution Mode implements a code execution pattern that exposes a single `run_python` tool to LLMs instead of hundreds of individual MCP tools, allowing the model to write Python code to discover and compose other tools dynamically. By reducing context consumption from 30,000 tokens to approximately 200 tokens per query, it provides a more efficient alternative to traditional tool exposure while supporting native data science libraries and offering enterprise-grade security through containerized, rootless execution. The server acts as a bridge that proxies any stdio-based MCP server, enabling zero-context tool discovery while maintaining isolation and security constraints.