Understand AI quotas and cost accounting

Use the context and quota indicators in Complexity chat to inspect your available allowance. An administrator controls tenant and group limits through AI governance.

The daily token quota belongs to each person. The most permissive applicable group limit wins. Unlimited daily tokens do not remove model or tenant context limits. Daily accounting uses the tenant's time zone.

The system can reserve tokens before a call. A reservation is temporary. It releases when the operation settles.

Run outcome	Personal token quota
Successful	Confirmed usage is charged.
Failed or interrupted	No final token charge.
Manually stopped	Confirmed usage through Stop is charged.

Actual provider cost can remain after failure. Administration therefore separates provider cost from personal quota charges. Missing usage does not mean zero consumption.

One accepted human message consumes one message allowance. Tool calls inside that run are not new human messages. Retry and Continue remain separate traceable operations. A successfully retried provider call within a run contributes to the complete operation's cost. See usage analytics for known and incomplete cost totals.