These examples do not install libraries or wrap application functions. Each one uses the same deployment pattern: route model traffic through Permisyn and attach authorization context as headers.
Risk: A lending agent can recommend denial or disbursement before a human sees it.
Control: Restrict the agent's passport to approved models and block calls once the daily cap is reached.
X-Permisyn-Agent: finance-report-agent X-Permisyn-User: finance-owner@company.com X-Permisyn-Team: finance X-Permisyn-Max-Cost-USD: 2.00
Risk: A triage agent routes urgent cases directly to operational queues.
Control: Force user and team attribution and purpose logging for every emergency-routing call.
X-Permisyn-Agent: triage-router X-Permisyn-User: clinical-ops@clinic.com X-Permisyn-Team: clinical-ops X-Permisyn-Purpose: patient triage
Risk: A contract agent can send redlines containing liability or IP changes.
Control: Restrict the agent's passport to an approved model/provider so out-of-scope calls are blocked before upstream.
X-Permisyn-Agent: legal-contract-agent X-Permisyn-User: legal-owner@company.com X-Permisyn-Team: legal
Risk: A support bot can issue credits, refunds, or account changes at high volume.
Control: Meter every stream, cap spend per agent, and generate signed evidence for escalations.
X-Permisyn-Agent: support-ticket-agent X-Permisyn-Team: support X-Permisyn-Max-Cost-USD: 0.25