Governance

Path: AI Foundation > AI Agent > Governance What it shows: the limits that apply to AI agent turns, and where each limit is configured.

Control	Effect on the AI agent
Monthly credit cap	Blocks turns when the group’s monthly credits are used up.
Daily request cap	Blocks turns after the daily limit per backend user.
Provider allowlist	Rejects turns that request a provider that is not allowed.
Model override	Rejects a requested model when model override is not permitted.
PII masking	Redacts email addresses and phone numbers in answers.
Turn guard	Built into the AI agent: warns at 20 tool calls and stops the turn at 40 tool calls.

What you do here: the tab is read-only. To change a limit:

Open AI Foundation > AI Permissions.
Select the backend user group.
In step Limits, set the caps, the provider allowlist, model override, and PII masking.
Under AI Agent tools, optionally enable Read-only agent (answers questions, prepares no changes) or enter Blocked tools as a comma-separated list, for example file_delete,cache_clear. The wildcard * is supported.
Save the group.

When a user belongs to several groups, the strictest setting applies. See AI Permissions.