mcp-use — Fullstack SDK for MCP Servers & Apps

Main

• Use the provided server SDK to define tools with explicit schemas; keep tool names short to avoid client-side limits.

• Treat /inspector as your dev loop: validate tool inputs/outputs before wiring the server into Claude/Cursor/ChatGPT.

• If you ship widgets, isolate UI code under a resources/ folder and keep tool logic separate from presentation.

• Standardize ports/transports per environment (dev vs prod) so clients always know where to connect.

README (excerpt)

 

About

mcp-use is the fullstack MCP framework to build MCP Apps for ChatGPT / Claude & MCP Servers for AI Agents.

• Build with mcp-use SDK (ts | py): MCP Servers and MCP Apps

Source-backed notes

• README shows a Python quickstart with pip install mcp-use and server.run(transport="streamable-http", port=8000).
• README states the inspector is available at /inspector for servers started with the built-in runner/listen flow.
• README includes a TypeScript example using MCPServer and server.listen(3000) with an inspector URL.

FAQ

• Do I need TypeScript to use it?: No—README includes Python quickstarts as well; pick one stack per project.
• Where is the inspector?: README examples show /inspector on the server port for local debugging.
• What should I deploy to production?: Deploy the MCP server endpoint; keep inspector access restricted to trusted environments.