We just launched no-code workflows on MCPCore. What would make you trust them in production?

Hi everyone, Halil here, the maker of MCPCore.

When we launched here eight months ago, MCPCore hosted your MCP servers, but you still wrote every tool in JavaScript. The feedback kept pointing at the same thing: hosting was solved, but writing the tools was still the hard part. Most people who want their AI assistant to reach a real API or database are not going to write a server.

So that is what we changed, and today it went live:

- Workflow builder: connect steps like HTTP Request, Database, Transform and Condition on a canvas, and it becomes a live MCP tool.

- OpenAPI import: paste a link or upload a file, then pick your endpoints. Read-only ones are selected for you. Anything that changes or deletes data stays off until you add it.

- AI drafts: describe a tool in plain words and get an ordinary workflow you can read and change.

- Failure handling: retries, error branches and For Each loops.

- Playground: chat with Claude or GPT using your own key and see every tool call with its parameters and result.

- Versioning: every save is a version. Work on a draft while the live version keeps serving, compare the two side by side and restore in one click.

- Clients: one click for Cursor and VS Code, one command for Claude Code, Codex and Gemini CLI, and it also works in ChatGPT as a custom MCP server.

I would really like your take on a few things:

1. If you have connected AI to a real API or database, what worried you most?

2. What would you need to see before trusting a no-code tool in production?

3. Which step or integration is missing for your use case?

The launch is live here if you want to take a look, and I will be around all day to answer questions. Honest criticism is very welcome:

https://www.producthunt.com/products/mcpcore