Show HN: Create MCP Servers and Agents Easily with PolyMCP (Python and TS) (github.com)

PolyMCP — a toolkit for exposing and orchestrating tools via the Model Context Protocol (MCP) — just gained a TypeScript-first implementation so everything you can build with the Python library is now available for Node.js/TS projects. The TS version brings full typings, Zod-based runtime validation, a minimal tool-definition API, and the same multi-server model (HTTP, stdio, and a new in-process mode). It also includes built-in API key/JWT authentication, a CLI for scaffolding and managing MCP servers/agents, Playwright stdio integration for browser automation, and a sandboxed executor for safely running LLM-generated code. For practitioners this matters: PolyMCP centralizes tool discovery and orchestration across HTTP and stdio MCP servers and plugs into major LLM providers (OpenAI, Anthropic, Ollama, etc.). The new CodeModeAgent compiles multi-step workflows into single code generations (claiming ~60% faster runs and ~68% less token use versus multiple tool calls), while the in-process server eliminates network overhead (40–60% speedups) for embedded agents. Additional capabilities — persistent memory, autonomous planning loops, mixed server support, and CLI-based registries — make it easier to build efficient, auditable, and production-ready agent systems that safely combine LLM reasoning with deterministic tools.