OpenAI SDK
The API speaks the OpenAI format. Keep your SDK and change the base URL.
Models, source control, deploy targets and payments. Connect a provider once in Connections: tokens are checked with the provider before they're saved, then stored encrypted.
The API speaks the OpenAI format. Keep your SDK and change the base URL.
Free Llama models through OpenRouter, when the platform has a key.
The 7B code model on NUCALORIC hardware, for builds, reviews and the terminal.
The 12B general model for writing, planning, reasoning and vision.
Free Mistral models show up in your model list when OpenRouter serves them.
Free DeepSeek models through OpenRouter, billed at the cloud rate.
Every local model runs on Ollama, so any model pulled on the host can be served.
Code comes in from GitHub, models from Ollama and OpenRouter, money from Stripe and Solana Pay. Sites go out to wherever you host.
Connect your tools once and every project in the workspace can use them.
Open models on NUCALORIC hardware. Gemma and Qwen Coder today, and any model pulled on the host.
The default general model: writing, planning, reasoning and vision.
Fast code model, plus a 32k-context build for large files.
OpenAI-compatible API at /v1. Point the official Python or JavaScript SDK at your NUCALORIC host.
Claude Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1 through the Anthropic API, billed per token.
The newest GPT models through the OpenAI API, in chat, agents and /v1.
Gemini models through the Gemini API.
Grok models through the xAI API, including Grok Build for code.
Mistral Large, Medium, Small and Devstral.
DeepSeek V4 models at low per-token prices.
Open models on Cloudflare’s edge: Llama, gpt-oss, Gemma, Qwen Coder, Kimi and DeepSeek.
Free cloud models, plus labs we don’t have a direct key for.
Repo import, branch and folder selection, signed push webhooks. A token is needed for private repos only.
Production deployments with a Vercel access token.
Site creation and zip deploys with a personal access token.
Panel URL, client API key and server ID. Files are written through the client API.
Publish any site to Pages on your own Cloudflare account with an API token. The project is created for you.
Publish to the gh-pages branch of a repo we create on your account, with Pages switched on.
Export, then serve the folder with a two-line nginx Dockerfile.
Export the site, or connect the same GitHub repo in Railway.
Export the site, or connect the same GitHub repo in Render.
Ship the exported folder in a small nginx image.
Accounts and workspace data in managed PostgreSQL. Any Postgres works, including our own servers.
Hosted site files stored durably in R2, cached on each server.
Bot protection on sign-up and sign-in, with no puzzles for people.
Sign in with GitHub. Accounts link by verified email.
Password-reset emails from our own domain.
Card top-ups through Stripe Checkout, credited from the signed webhook.
USDC top-ups, verified on-chain by reference. Solana is used for payments only.
Repo import and push webhooks, like GitHub.
Static sites on App Platform.
Cloud Storage and Firebase Hosting.
Nothing matches that yet. Tell us what to connect next.