Hosted runtime for portable, WASM-sandboxed MCP servers ('servlets') that can be plugged into Claude, ChatGPT, Cursor, and custom agents.
Hosted runtime for portable, WASM-sandboxed MCP servers ('servlets') that can be plugged into Claude, ChatGPT, Cursor, and custom agents.
mcp.run is a managed runtime and registry for Model Context Protocol servers, branded on the homepage as 'Turbo MCP.' Instead of asking developers to clone a GitHub repo and run an MCP server locally, mcp.run packages each server as a portable WebAssembly 'servlet' that runs in its sandbox and is callable from any MCP-aware client over a single hosted endpoint. That means you can install dozens of MCP tools — search, scrapers, databases, internal APIs — and your laptop never spins up a Node or Python subprocess per server. The platform handles auth, secrets, per-tenant isolation, and updates, and offers a content registry where developers publish new servlets that others can subscribe to with one click. Because servlets are WASM, the same server runs identically on a developer's machine, in a CI pipeline, and inside the hosted runtime. mcp.run is targeted at teams that want MCP as their integration substrate but don't want to operate a fleet of MCP processes per user. The site is heavily JavaScript-rendered, so specific pricing tiers and limits need manual verification.
Contact / see site
Ready to get started with mcp.run (Turbo MCP)?
View Pricing Options →