Fastmcpp (Fastmcp for C++) (github.com)

fastmcpp is a new high-performance C++ port of the Model Context Protocol (MCP) intended for building native MCP servers and clients with low overhead. Announced as beta (v2.13.0), it implements MCP’s core JSON‑RPC behavior (following the Python fastmcp as the canonical reference) and adds first‑class support for tools, resources, prompts, middleware, and JSON Schema validation. Multiple transport layers are supported out‑of‑the‑box — STDIO, HTTP (SSE) and WebSocket — making it suitable for embedded agents, local CLIs, and networked model-serving integrations where C++ performance and predictable resource usage matter. Technically, fastmcpp targets C++17+ with CMake 3.20+, is cross‑platform (Windows/Linux/macOS), and uses nlohmann/json (auto-fetched). Optional components include libcurl for HTTP POST streaming and bundled cpp-httplib / easywsclient for server/client functionality. The repo includes example servers/clients (stdio_server, server_quickstart, client_quickstart), a GoogleTest suite (24/24 non‑streaming tests passing; some streaming tests disabled for infra reasons), and CMake flags to enable post‑streaming and streaming tests. Apache‑2.0 licensed and community‑driven, fastmcpp is useful for ML/AI projects that need a lightweight, native MCP stack (tool invocation, middleware hooks, prompt/resource management) with easy integration into existing C++ systems while tracking behavior parity with the established Python implementation.