Making Agents Useful: MCP, Tool Calling, and Sandboxed Execution

FastMCP provides an ergonomic Python framework for building Model Context Protocol (MCP) servers, enabling agents to interact with external tools. Traditional MCP implementations suffer from "context bloat," where injecting extensive tool schemas into an LLM's context window degrades performance and increases latency due to serial execution. "Code Mode" mitigates these issues by allowing agents to search for tools and execute arbitrary, secure programs within a Modal sandbox. This approach replaces dozens of individual, serial tool calls with a single program execution, significantly reducing context consumption and latency. By offloading untrusted code execution to isolated sandboxes, developers maintain a consistent, performant user experience while scaling their toolsets. This architecture effectively separates the control plane from the data plane, ensuring that agents remain responsive even when managing complex, multi-step workflows.