Provider keys and custom models for chat and the builder. Each provider card shows where its key comes from — a house key on the server, your own .env, or this browser. Keys you add stay in your browser. Appearance, memory, and usage live in the chat settings popup.
Ultra-fast inference for chat and small builder edits.
Custom models
gpt-oss-120b is the default Cerebras model and the engine behind the Jomoq 1 fast pass. It runs on the house key out of the box; add your own key below to use your own quota instead.
API keys
Access many models through one API key.
Custom models
Paste any OpenRouter model ID, then add it to the model dropdown.
API keys
Free keys at build.nvidia.com. Presets appear in the model menu once added.
Fast open-model inference for chat and the builder. Free keys at console.groq.com/keys.
Deepseek models for chat and the builder. Keys at platform.deepseek.com/api_keys.
Fast Apply and related tools. Keys at morphllm.com/dashboard/api-keys.
Primary quick web search for chat and builder: fast snippets and AI summary. Keys at app.tavily.com.
Deep research for finance (earnings, SEC/IR scrape) and academic, GitHub, or PDF sources when web_search uses those categories. Keys at firecrawl.dev.
Caps total web_search tool calls per assistant reply across both providers (default 5). Quick searches use Tavily; deep research uses Firecrawl scrape.
Live public data sources the AI can query. Most work out of the box; a few take an optional key.