MCP Playground
A Python toolkit and interactive Streamlit playground for talking to remote Model Context Protocol servers over Streamable HTTP or SSE, with LLM-powered tool selection across several providers.
MCP Playground provides an async Python client for connecting to a remote MCP endpoint, listing its available tools and calling them, supporting both the default Streamable HTTP transport and the older SSE transport for legacy servers, plus custom authentication headers through a supplied HTTP client. A bridge layer lets a chosen language model pick and execute the right MCP tool for a natural-language request, with built-in bridge classes for OpenAI, Anthropic, Ollama and OpenRouter.
A companion Streamlit application turns the same client into an interactive playground with a transport selector, multi-provider model choice, chat-only, tools-only or automatic chat modes, tool discovery and invocation, and real-time connection status, useful for exploring or debugging an MCP server without writing code.