Parse
Drop in a spec file or URL. Every operation becomes a typed tool with a schema an AI can read.
Upload an OpenAPI spec or Postman collection, choose the endpoints, and get a hosted MCP server your AI clients can call. Your API keys stay encrypted and never leave the gateway.
01 · Engine
From spec to server in four steps.
No code to write and nothing to host. Upload, choose, connect: the gateway handles schemas, auth, timeouts and errors the same way for every client.
Curate
Switch on the endpoints that matter, and rename them so the model picks the right one.
Serve
One MCP URL per server. The gateway turns each tool call into the real HTTP request.
Protect
Upstream keys are encrypted and write-only. Private addresses are blocked on every call.
02 · Selection
Expose the endpoints you choose. Leave the rest dark.
A spec can hold hundreds of operations; a model works best with a focused few. Turn on the ones that matter, and the rest never appear in the tool list.
72 operations · 20 exposed
03 · Numbers
Built for real APIs, with real keys.
Big specs parse in seconds, every call is logged, and the credentials your API needs stay yours.
2,000operations read from a single spec
100active tools per server, each one chosen
0API secrets readable after you save them
7 daysof request logs, with secrets redacted
04 · Questions
Before you connect.
The short answers. The rest is in the dashboard, next to the setting it's about.
Which AI clients can connect?
Anything that speaks MCP over HTTP. We test Claude (web, desktop and Claude Code), Cursor, VS Code and ChatGPT. Clients sign in with OAuth, or send an API key in the Authorization header.
What happens to my API keys?
They're encrypted the moment you save them and only decrypted in memory for the call that needs them. No screen or API ever shows them again, and they never reach logs, caches or the AI client.
Can the AI call every endpoint in my spec?
Only the ones you switch on. Calls that can change data are marked as such for the client, and the gateway only ever calls your API's own base URL.
What does it cost?
Nothing during the beta. Every account has the same limits, listed on the limits page. See the limits
