Credits and limits
Understand what connected-assistant runs cost, why runs use a quote, and how the integration spend limit works.
Runs started through Claude, ChatGPT, or Claude Code use the same organisation-wide credit balance as runs started in the web app.
Quality Check
A Quality Check costs 2 credits per run on every plan. The assistant must show and confirm the credit cost before starting it.
Simulation
A Simulation costs 40 credits. The assistant first requests a quote for the exact document revision and persona. The quote is valid for five minutes and does not charge or reserve credits. On Free, the quote is refused because Free does not include Simulation.
Starting the Simulation is a separate action. You approve that action in your assistant, and the run starts only if the quote is still valid and the organisation can cover it. If the quote expires or the document changes, request a new quote.
The spend window
Your organisation also has a shared limit for credits committed through connected assistants. The default is 10,000 credits per calendar month. Owners and admins can change it in Settings → Integrations.
The limit counts credits when a run is accepted, so it can be reached before every run has finished and settled. It is shared by everyone in the organisation across Claude, ChatGPT, and Claude Code. Runs started in the web app do not count toward this integration limit.
When the limit is reached, the next paid run is refused even if the organisation still has credits. The refusal tells you when the calendar-month window resets. You can also ask your assistant for the workspace context to check the remaining room before a batch.