Model Context Protocol (MCP) has become the default way to give AI assistants access to external tools and data, and now there is a cloud built just for it. Manufact, the startup behind the open-source mcp-use framework, has launched Manufact Cloud, a hosting platform that takes an MCP project from a GitHub push to a live, monitored endpoint that Claude, ChatGPT, Gemini, and Copilot can all call. The company introduced it in a Launch HN post on July 2.
How to Integrate It
Connect a GitHub repo and every push auto-deploys to a live MCP URL, with branch previews for testing before you promote a change. The bundled Cloud Inspector lets you fire tool calls, inspect the raw JSON-RPC traffic, and swap models to debug a server while it runs. If you are building a custom connector, say an MCP server that exposes your asset library or content pipeline to Claude, you get hosting, authentication, and analytics without standing up your own infrastructure. The mcp-use SDK it builds on ships for both TypeScript and Python, and a single codebase can deploy to ChatGPT, Claude, Gemini, and Copilot at once.
Why It Matters for Creators
Shipping an MCP server to production has meant hand-rolling hosting, auth, and observability, then fighting store review to get listed. Manufact folds that into one workflow and generates marketplace checklists and submission assets for the ChatGPT Apps Store, Claude Connectors, and Gemini Enterprise. As the founders describe in their Launch HN post, the goal is a single lifecycle: scaffold, inspect, deploy, publish, and monitor. The mcp-use framework it is built on already has more than 10,000 GitHub stars and 7 million downloads, so the launch arrives with an established builder base rather than from a standing start.
Key Details
Company: Manufact, part of the Y Combinator S25 batch, based in San Francisco and Zurich.
Two parts: the open-source mcp-use SDK (TypeScript and Python) and Manufact Cloud hosting.
Features: zero-config GitHub deploys, Cloud Inspector debugging, branch previews, built-in auth, and usage and latency analytics.
Pricing: usage-based, with free credits for trial deployments.
What to Do Next
Scaffold a server with the mcp-use SDK, connect your GitHub repo, and push to get a live MCP URL you can test inside Claude or ChatGPT. Builders already running MCP servers can point the Cloud Inspector at production traffic to trace and replay tool calls before submitting to a marketplace, then use the generated checklist to clear store review.
