This project implements Anthropic’s Programmatic Tool Calling (PTC) paradigm, allowing AI agents to invoke tools by writing and executing code rather than making individual JSON tool calls. Built on LangChain and LangGraph, it uses the Model Context Protocol (MCP) to enable agents to orchestrate entire workflows through generated code. The approach reduces context window pollution by letting models write self-contained code that handles data fetching and processing internally, then returning only results.