The same CAD kernels that run in our browser editors, available to your backend over plain HTTP. Generate parametric models from parameters, convert STEP files to glTF, and chain kernel operations into pipelines - from Node, Python, Go, Java, .NET or anything else that can make a request.
Running a geometry kernel in the browser is the right answer for an interactive configurator, and the wrong answer for a nightly batch of ten thousand parts, a quoting service that must not trust the client, or a pipeline that has to produce a manufacturing-grade STEP file. Those belong on a server, and until now that meant licensing a desktop CAD kernel and operating it yourself.
The CAD Cloud API is that server. You send a request, it runs on managed compute, and you download the result in the format you need.
occt.shapes.solid.createSphere.Every request is asynchronous by design, because real CAD work is not measured in milliseconds. You submit a job, receive a task id, poll for completion, and download the result as STEP, glTF, GLB, a decomposed mesh or one of the other supported formats. That model holds whether the job takes two seconds or two minutes, so a slow operation never becomes a timed-out HTTP request.
Authentication is an x-api-key header on every endpoint except the health check. Keys are scoped, so a key issued for conversions cannot be used to run arbitrary kernel operations.
The whole surface is described by a published OpenAPI 3.1 specification, so you can generate a client in any language, import it into Postman or Insomnia, or hand it straight to an AI coding assistant. A first-party TypeScript SDK is on npm if you would rather not generate one.
The project CLI sets up a full-stack application with the API already wired in. Choose the CAD Cloud option and you get a backend - Hono on Cloudflare Workers, Node.js with Express, or ASP.NET Core - and a React and Three.js frontend that proxies requests through it, so your API key never reaches the browser.
npx @bitbybit-dev/create-app my-project