Included model usage is measured by the work the agent does, not by a fixed message count. Long sessions, large schemas and complex pipelines use more. Pro sets the baseline, Max gives 5x or 20x more, and limits apply to each person, so one heavy user does not reduce what teammates have. The exact allowances are not set yet.
When you reach a limit you have four options: wait for it to reset, switch the agent to your own model key, move to a plan with more included usage, or turn on usage credits, which keep the agent working at standard metered rates until you turn them off or reach a spend limit you set.
Your projects, templates, local data and dashboards stay available. Work that does not need the model, such as golden tests on DuckDB, keeps working. See usage credits and the pricing page.