PROVIDER-NEUTRAL DESIGN COPILOT
Choose copy/paste with your preferred assistant or connect directly to an API and watch its response stream into the project.
Direct API mode: your browser sends the prompt (including the open project in Modify mode) and API key directly to this Base URL; they do not pass through Circuit Studio. The provider receives them. OpenAI recommends keeping API keys server-side. Remembered keys are stored unencrypted in this browser's local storage, where scripts running on this site can read them. Key memory is off by default; use a limited key. OpenAI key guidance
Why a proxy? Token verification works straight from the browser, but OpenCode's generation endpoints send no CORS headers and answer preflight with 404, so no static page can call them directly. If you started the app with node server.js, a local relay is used automatically and nothing leaves your machine except the call to OpenCode. Otherwise a site default or your own proxy (proxy/opencode-cors-worker.js) is used. Any proxy forwards your token, so it sees it: only use one you trust. Requests go browser → proxy → OpenCode.
node server.js
proxy/opencode-cors-worker.js
OpenCode Inference: Verify calls GET /inference/v1/models with your token. Generation sends the token and prompt to OpenCode's family-specific inference endpoint, through a CORS proxy when one is used; it does not go through Circuit Studio. Free OpenCode models are reserved for the OpenCode app, so use a paid model with Console credits. If remembered, the token is stored unencrypted in this browser's local storage. OpenCode's token and endpoint docs: Inference API.
GET /inference/v1/models