No JSON schemas. No infrastructure. Connect your existing API and any AI agent can call it — in minutes, not days.
No credit card required · Up and running in 10 minutes
No new code to write. No schemas to maintain. Just point us at your existing API.
Import directly from GitHub or paste your controller file. We support Spring Boot today — Express, FastAPI, and Rails are coming soon.
Claude reads your endpoints and writes rich, semantic tool descriptions — the kind that tell an AI agent exactly when and why to call each one.
You get a hosted MCP URL. Paste it into Claude Desktop, Cursor, or any MCP-compatible client. Your AI can now call your real API.
No sign up needed to preview. Paste your code or import from GitHub and watch the magic happen.
Needs contents:read scope. Never stored.
Scanning your code...
Your server lives at a stable URL — mcpforge.io/t/your-slug/mcp. No servers to manage, no uptime to worry about.
We don't just expose your endpoints — we write descriptions that help AI agents understand intent, not just schema. Better descriptions = smarter agents.
Each MCP server gets a unique API key. Only clients with your key can invoke your tools — your backend stays protected.
See which tools get called, how often, and whether they succeed. Know exactly how AI agents are using your API.
Connect your repo and your MCP server updates automatically when you push. Your tools stay in sync with your code.
Built for developers who want to ship, not configure.
From zero to a working MCP server in under 10 minutes. Paste your controller, enter your API base URL, deploy.
Claude reads your code and writes MCP tool definitions with semantic descriptions — no manual JSON schema authoring.
We run the MCP server. You never touch a Dockerfile, a load balancer, or an SSL certificate.
Push to your repo, we rebuild your tools. No manual redeploys. Your MCP server tracks your codebase automatically.
Real-time analytics on tool calls, success rates, and monthly usage. Know how your API is being used by AI.
Bearer token authentication on every endpoint. Your backend URL is never exposed to MCP clients.
If you have an HTTP API and want AI agents to use it, MCP Forge is for you.
You already built the API. Don't rewrite it as MCP tools by hand — point us at your controllers and we handle the schema generation and hosting.
Your agent can call real endpoints — fetch data, submit forms, trigger workflows — without you writing a single line of MCP glue code.
Internal APIs become AI-accessible overnight. Customer support, ops, data — anyone with Claude Desktop can query your systems through natural language.
Start free. One MCP server, 500 tool calls a month. No credit card. Perfect for experimenting and shipping fast.
Model Context Protocol (MCP) is an open standard that lets AI assistants like Claude call external tools and APIs. It's how you give an AI agent the ability to act, not just talk.
No. We sit in front of your existing API. Your backend doesn't know MCP exists — it just receives regular HTTP requests from us.
Spring Boot is fully supported today. Express, FastAPI, and Rails are in active development. If you have a specific framework you need, email us.
No. We process your controller file in memory to generate the MCP tool manifest, then discard it. Only the generated tool definitions are stored.
Every time an AI agent invokes one of your tools — e.g. Claude calling your getPayment endpoint. Initializing the server or listing tools is free.
Yes. Redeploy any time from your dashboard by uploading new code, or enable GitHub auto-sync to update automatically on every push.
Free to start. No credit card. No configuration headaches.
Get my free MCP server →Free plan · 1 MCP server · 500 calls/month · No credit card