Dashboard, Budgets and Alerts

Part of Monitor.

The Monitor Page

• Overview: spend against the period before, tokens, cache hit rate and savings, month to date with a forecast, estimated active time, error-free share, plan value against what you pay, spend per day by agent, where tokens go, and the plan windows closest to their limits.
• Limits: every plan window with its projection, run-out time, the use line for the current cycle and earlier peaks.
• Models: per model: responses, tokens, spend and share, cache hit rate, cost per million tokens, reasoning share, response time, peak context and errors; effort, reliability, and a weekday-by-hour spend heatmap.
• Projects: spend, sessions and active time per repository, and the costliest sessions.
• Team (Max and Ultra): each member’s usage for owners and admins, and every machine with what its collectors found.
• Settings: how long your plan keeps Monitor data and when the oldest ages out, the organization’s monthly budget, what you pay for each plan, and a CSV export.

Pick a range above the views (ranges longer than your plan keeps are greyed out) and filter by member, agent and project; the view updates as you choose. Ask your agent holds the prompts for your coding agent. Owners and admins on Max and Ultra see every member; everyone else sees their own usage.

Budgets and Alerts

• Plan limit alerts email you when one of your plans reaches 90% of a window, and when it runs out, once per window each cycle.
• Budget alerts (Max and Ultra) email owners and admins when the organization’s spend this month reaches each percentage of its budget, checked each night. Under Settings, enter the budget, tick up to five percentages (or type your own under Custom %), and the page shows which ones today’s pace reaches this month.

Turn either off on your Profile and security page or with the link in the email.