Prove what your AI agent was allowed to do.
Every agent gets a cryptographically signed passport. Every sensitive action is verified before it runs. Every decision lands in a tamper-evident audit trail.
Free tier, 10,000 verifications/month. No credit card.
Same agent, two requests. Real responses from the live API.
Try it right now
A real agent, a real decision, no signup.
CustomerSupportAgent · production permitted: read:customer_records, send:support_response denied: write:billing_records, delete:customer_data
That decision was just written to a tamper-evident audit log. Start free to see yours
Logs and IAM don't answer this.
Logs record events, not authority
They can't tell you what an agent was permitted to do, they're written after the action, and whoever owns the database can edit them.
IAM authenticates services, not agents
A service account proves the application is allowed in. It says nothing about what the agent inside it may do, on this action, right now.
Auditors are already asking
The EU AI Act is in enforcement. “What did your AI do with our data, and was it authorized?” is becoming a procurement question.
How it integrates
const decision = await veris.verify({
passportId,
requestedCapabilities: ['read:customer_records'],
});
if (decision.decision === 'ALLOW') { /* proceed */ }Wire it into your deployment pipeline once — under a day. Every agent after that gets a passport automatically. Trust scoring and policy details live in the docs.
Start freeWhere we are
The protocol, signed passports, policy engine, and audit chain are live and in use today. HSM-backed signing, independent penetration testing, and SOC 2 are on the near-term roadmap — ready for pilots with non-sensitive data now. Our threat model is public, including what isn't hardened yet.
Read the threat model →Start verifying in minutes
Free tier includes 10,000 verifications per month. No credit card required.