Choose which models your workspace and Auto can use

Settings, Models now shows your agent models as a live map of where they come from. Three sources sit on the left: models General Input hosts, members' Claude, ChatGPT, Gemini, Grok, and Muse subscriptions, and your custom endpoints. Every model sits in the middle with its switch, and lines run on into your workspace. Use it for cost control: keep a team on Auto and a few affordable models.

How it works

• One switch per model, across every source. Turn a model off and nobody in the workspace can run it, whether from General Input, a member's subscription, or a custom endpoint. Another turns off General Input's hosted models entirely, keeping subscriptions and custom models.
• See what's live. A green line shows every way a model runs now; a model nothing serves says so.
• Click a source to use it. Subscriptions opens to connect your accounts, share them with the workspace, and turn your own models off. Custom opens to add or edit your endpoints' models.
• Everyone sees it, admins change it. Every member sees the map and can connect and share subscriptions; only admins switch models.
• Auto follows your switches. Auto tries connected subscriptions first, then hosted models, running only models that are on. A workflow set to a model that is off runs on Auto; a chat moves to the member's default or Auto.

Also shipped

• Click a model for its card: its credits per 1K tokens, context window, and what it reads. Each provider lists its priciest first.
• The utility model gets its own section: pick the one small model that names chats and handles other background work.
• The "Default to subscription models" switch is gone: turn a model off to keep Auto off it.