What it does
AI works on every plan without connecting anything — requests go through our own AI provider accounts. Connecting your own Anthropic key is optional: it moves the work you route to Anthropic onto your own Anthropic account, under your own agreement with Anthropic.
Once connected, Claude models are one of the choices when you decide which model handles which kind of work, alongside OpenAI and Google Gemini if you connect them too.
Setup steps
1. In Anthropic, create an API key.
2. In your AI provider settings, connect Anthropic and paste the key. A workspace admin can do this; it does not need us. The key is encrypted at rest and only ever shown masked afterwards.
3. Test the connection from the same screen.
4. In AI routing, choose which kinds of work go to Anthropic and which model handles them — one rule for a single use, or one for a whole family of uses.
5. Optionally, set budgets in AI budgets — a cap in dollars per use and period. Calls that would go over the cap are refused until the period resets, rather than billed.
What you get
- Your account, your terms. Work routed to Anthropic runs on your Anthropic account.
- One set of rules. Whichever model does the work, the AI uses the same actions and permissions as the person who asked, and is never given anything irreversible to do.
- Every call accounted for. The AI audit page shows which provider and model handled each call, for which use, and what it cost.
Related
See the AI page for how the AI works across the product, and the sub-processor list for the AI providers that handle your data.