Ternivo Connect · Settings

AI agent access

Control exactly what Claude or another OAuth AI client may do. Ternivo remains the enforcement point even when the AI host can call write tools.

Back to Settings

Organization guardrails

Detected AI clients

Clients appear after they connect to Ternivo MCP. Configure one to grant more than the safe read-only default.

Active and historical grants

Configure an AI client

Default deny: without an active grant, OAuth AI clients receive only safe connection/delivery/diagnostic/analytics reads. Public publishing still requires the organization-level policy to permit it.

OAuth client ID

Label

Expires

Tool permissions

Publishing destinations

Only write-enabled accounts can be granted for publishing. Read-only bridged accounts remain visible to the agent but cannot be used for writes.

Allowed media asset IDs (optional, comma-separated)

Paid-media activation limits

Only used when the high-risk paid activation scope is granted. Claude can never approve its own spend request.

Maximum daily budget

Maximum lifetime budget

Private/test publish

Allows private/test delivery when the selected provider supports it.

Public publish

Still ineffective while organization-level agent publishing is locked off.

Scheduling

Requires both the schedule scope and organization scheduling policy.