Which Claude plans work?

FAQ. What Cardinal needs from your Claude account, and how Desktop, Cowork and API keys fit in.

Cardinal works with your own Claude account. It never asks for your Claude login and never uses a key of its own.

For AI drafts (the runner)

The runner uses the Claude Code CLI on your computer, signed in with your Claude plan. The installer checks claude auth status and says "signed in with a Claude plan" when it is.

Signed in with an API key instead: the installer and the runner's check warn you. Run claude auth login to switch to your plan.
Claude Code older than 2.1.268: the installer asks you to run claude update.
No Claude Code CLI: Cardinal can still count your use (for example from Claude Desktop), but it can't draft with AI on that computer.

Cardinal asks for model aliases, sonnet for longer writing and haiku for short tasks, so they follow what your plan offers. AI jobs use your plan's own allowance.

For counting your use (telemetry)

Claude Code CLI: counted.
Claude Desktop's Code tab: counted; it reads the same settings file.
Cowork: doesn't read these settings on individual plans. On Claude Team or Enterprise, your admin turns on Cowork telemetry in the admin console instead.
Chats on the Claude website or apps: not counted.

Seats on the Team page

Execs can record each person's seat as Pro, Max 5×, Max 20× or Team, with its price. Cardinal uses that to work out the value ratio (API-equivalent against seat price) and idle seat cost. It doesn't check your plan with Anthropic.

API-equivalent is not a bill

API-equivalent is what the same usage would cost at API prices, as Claude Code reports it. It is a measure of value, not what Anthropic charges you.