Forge Mcp Server

Forge MCP Server is an MCP (Model Context Protocol) server that enables AI coding assistants—including Claude, Cursor, Windsurf, OpenCode, and VS Code Copilot—to convert PyTorch models into highly optimized CUDA/Triton GPU kernels. It exposes tools for authenticating to the Forge service, submitting PyTorch code for automated kernel optimization, and generating new GPU kernels from natural language or specific technical requirements. Utilizing 32 parallel multi-agent swarms, Forge benchmarks and validates kernels on real datacenter GPUs (from T4 to B200) for robust, production-ready speedup (up to 14x). It is ideal for machine learning engineers, researchers, and developers seeking fast, production-grade GPU code for their AI workloads.