Launching today
MCPCore
Build AI-powered MCP servers in the cloud
38 followers
Build AI-powered MCP servers in the cloud
38 followers
Define your AI tools, deploy in one click, get a live endpoint with its own subdomain instantly. Build unlimited tools with the AI-powered tool builder. Bring your own custom domain. Flexible auth from public access and API keys to full OAuth 2.0 with PKCE. Built-in rate limiting, encrypted secret management, real-time traffic logs, and error tracking out of the box. No servers, no infrastructure, no ops. Start building for free
MCPCore
Hey everyone! I'm Halil, the maker behind MCPCore.
I built this because I got tired of the same setup every time I wanted to connect an AI tool to my own stuff. You want Claude to query your database or Cursor to hit your internal API and suddenly you're writing a whole backend, setting up auth, deploying, configuring domains. It felt like way too much effort for something that should take minutes.
MCPCore handles all of that. You open the browser, write your tool logic in the editor, hit deploy, and you get a working endpoint right away with its own subdomain.
The cool part is what you can actually build with it. Give Claude access to your product catalog so it can answer customer questions from real data. Let Cursor query your database while you code. Connect Windsurf to your internal APIs so it can pull reports for you. Hook up any AI assistant to your CRM, your inventory system, your analytics dashboard, whatever you need. If it has an API, you can make it available as an MCP tool in a few minutes.
For those who don't want to write code at all, there's an AI Builder. You describe what you want your tool to do in plain English, and it generates the handler logic for you. Just review it, hit deploy, done.
Everything runs in isolated environments so your code doesn't touch anyone else's. Secrets and credentials are stored with strong encryption. Auth is flexible, you can keep it open, lock it with an API key, or go full OAuth.
There's a free tier so you can start building right away. Would really appreciate your honest thoughts.
this is exactly what I've been wanting. setting up MCP servers manually is such a pain, especially the auth and deployment part. the browser based editor is a great call too since most devs already live in the browser anyway. curious if you support custom middleware or if the tool logic has to be pretty simple right now?
MCPCore
@mihir_kanzariya Thanks! Yeah the auth and deployment part is exactly what drove me crazy too, glad I'm not the only one haha.
For the tool logic, you're not limited to simple stuff at all. The editor runs in a sandboxed JavaScript environment with built-in SDK support for HTTP requests and database queries, plus lodash is available out of the box. There's also an AI Builder if you want to generate handler logic from a plain English description.
As for middleware, there isn't a separate middleware layer but you can handle that kind of thing inside your tool logic. For example you can validate the incoming request at the top of your handler, check conditions, and return early with a message back to the AI if something doesn't look right. So you could do things like "if this parameter is missing return an error to the LLM" or "if this user isn't authorized tell the AI the request is invalid" before running your actual logic. Rate limiting and logging are already built in on the platform side, you can check your logs from the Logs menu in real time.
What's on your mind, what would you build first? I'm shaping the roadmap around real use cases so this kind of input really helps.