Every AI agent in your org, accounted for
CISOs lose sleep over three questions: Which agents are running? What can they access? Who owns the incident? Pretense answers all three: enforcement, not just visibility.
The 80-point enterprise readiness gap: answered
Which agents are running?
Pretense auto-discovers every AI agent routing through your network. Enrolled agents are profiled; unregistered agents trigger an immediate alert.
What are they authorized to do?
Define per-agent policies: which codebases, which data classes, which actions. Pretense enforces at the request level, not just logs after the fact.
Who is accountable when one goes wrong?
Every agent has a named owner. Every incident auto-creates a Jira/Linear ticket with CISO escalation SLA. Audit trails are 7-year retention for SOC2.
Agent Inventory
All active AI agents (live view)
| Agent | Status | Risk | Mutations | Secrets Blocked | Owner | Last Active |
|---|---|---|---|---|---|---|
Claude Code (eng-team) Anthropic | Protected | low | 1,284 | 7 | Platform Eng | 2 min ago |
GitHub Copilot (IDE) OpenAI | Protected | low | 892 | 3 | DevEx | 8 min ago |
Cursor (product team) Anthropic / OpenAI | Scanning | medium | 341 | 1 | Product Eng | 23 min ago |
Unregistered Agent Unknown | Threat Detected | critical | 0 | 0 | Unassigned | 47 min ago |
Critical: Unregistered AI agent detected
An agent not enrolled in Pretense policy was detected making LLM API calls. All traffic is blocked until policy is assigned. Incident #INC-2847 created. Escalation sent to CISO.
Authorization Matrix
Define what each agent can and cannot touch
Traditional DLP tools log what happened. Pretense prevents it. Set per-agent policies at the data class level, and Pretense enforces them inline, before any request leaves your network.
- Per-agent policies: allow/block by data class (PII, secrets, IP)
- Per-repo rules: different sensitivity for frontend vs. payments code
- Time-bound access: temporary elevated permissions with auto-expire
- Emergency revoke: kill a compromised agent in < 60 seconds
Incident Accountability
When something goes wrong, who owns it?
Pre-defined escalation paths so no one scrambles when a breach happens. SOC2 requires documented accountability. Pretense builds it in.
| Incident Scenario | Primary Owner | Escalation Path | Response SLA |
|---|---|---|---|
| Agent leaks proprietary code to API | CISO | Legal + Engineering VP | < 15 min |
| Secret (API key) detected in prompt | Security Eng | CISO | < 5 min |
| Unregistered agent detected | Platform Eng | CISO | < 30 min |
| Agent authorization violation | Compliance | Legal | < 1 hour |
| High-volume data exfiltration attempt | Security SOC | CEO + Legal | Immediate |
Integrates with your existing security stack
Ready to answer your CISO audit?
Enterprise teams get full agent governance: inventory, authorization, accountability, SIEM integration in one deployment.
SOC2 Type II in progress • HIPAA ready • ITAR compatible • 7-year audit retention