Models overview
Command Code built-ins, custom OpenAI-compatible endpoints, and the composer picker.
Open /models to see what is available in chat.
Brain consumes OpenAI-compatible chat completions. It does not train models; fine-tuning and content filters belong at the model host (for example Azure AI Foundry). See Architecture.
Built-in (Command Code)
When COMMAND_CODE_API_KEY is set on the host, curated models appear as Built-in. Workspace owners/admins (or the instance admin) can turn each one on or off on /models, or use Turn all off / Turn all on. Off models are hidden from that workspace's composer. They are not deleted — turn them back on to restore them.
If no Command Code key and no custom models exist, chat shows an unavailable empty state until you configure one path.
Native Anthropic Messages (Claude-as-provider) is not in the curated picker — that API shape is different. Claude.ai / ChatGPT as the product staff open is also out of scope: Brain is the private client.
Custom (including Azure AI Foundry)
Instance or workspace admins can register any OpenAI-compatible base URL — Foundry, a company proxy, Ollama, LM Studio. Chat can run without COMMAND_CODE_API_KEY when a usable custom model exists for the workspace. See Custom models.
Composer picker
The chat composer remembers your last model. Choose a built-in or custom model.
Tools catalog
Connection tools live on /tools — see Tools catalog.