Trust. Verified.
Every AI Agent.
The identity and trust infrastructure for autonomous AI. Issue cryptographically verifiable AI Passports. Verify any agent in under 50ms.
- SOC 2 Ready
- <50ms p99
- Multi-cloud
How it works
From issuance to verification in three steps. No SDK lock-in.
Issue Passport
Your org registers each agent with a capability scope and trust baseline.
Embed in Agent
A single SDK call on init injects a signed JWT into every outbound request.
Verify Anywhere
Relying parties get a <50ms trust decision over standard HTTPS.
Trust evidence, and what you do with it
Veris standardizes five independently verifiable signal families and returns each one with every decision. How you weight them is your choice. The composite below is the protocol's reference model — sensible defaults so you can start immediately — and any organization can substitute its own weights or ignore the composite entirely and apply its own risk engine to the raw components.
Reference model — the protocol's default weighting. Configurable per organization.
T(a) = 0.30·I(a) + 0.25·B(a) + 0.20·C(a) + 0.15·H(a) + 0.10·E(a)
At issuance, Veris records a behavioral baseline hash derived from the agent's declared scope and operational profile. Drift is measured as statistical deviation from that baseline. Minor drift reduces the score. Sustained or severe drift triggers an automatic ATTESTATION event and alerts the organization.
| Drift Level | Score Impact | Audit Severity |
|---|---|---|
| None detected | 0 | INFO |
| Minor drift (< 2σ) | -30 | MEDIUM |
| Major drift (> 2σ) | -100 | HIGH |
| Scope violation | -200 | HIGH |
| Security incident | -500 | CRITICAL |
What the protocol standardizes
- The five signal components and how each is measured
- The signed passport format and its verification
- The capability scope model and precedence rules
- The audit event format and hash chain
- The verification request and response contract
- The weight of each signal in the composite
- The tier thresholds
- Policy rules layered on top of the score
- Whether to use the composite at all
Try the formula yourself
| Signal | Weight | Score | Contrib. |
|---|---|---|---|
| Identity | 30% | 920 | 276.0 |
| Behavioral | 25% | 960 | 240.0 |
| Compliance | 20% | 900 | 180.0 |
| Historical | 15% | 800 | 120.0 |
| Environmental | 10% | 850 | 85.0 |
| Composite | — | — | 901.0 |
This simulation uses the exact formula from the open source Veris Protocol spec. Weights are fixed by the protocol. Individual signal values are determined at verification time by the Veris verification engine.
What the score means
Default thresholds. Configurable per organization.
Low and Critical tier decisions trigger automatic alerts and are logged as HIGH severity audit events. Relying parties configure their own response policies.
→ View the full algorithm in the open source specHow trust is established — and what we don't claim
Veris Protocol operates on a layered trust model. When an organization registers an AI agent, they declare its capability scope. Veris signs and verifies that declaration cryptographically — but we do not independently verify whether the declared scope is accurate.
This is a deliberate design choice, not a gap. Veris establishes verifiable identity and enforces declared boundaries. The accountability for declaring accurate scopes sits with the issuing organization — the same way a company is accountable for what permissions they grant a human employee.
On standardizing risk. Veris deliberately does not impose a single risk model. Protocols that have achieved broad adoption standardize formats, identities, signatures, and verification methods — not reputation or risk scoring. TLS specifies how a certificate is structured and verified; it does not tell a browser which authorities to trust. Veris follows that pattern: we standardize what can be proven about an agent, and leave to each organization the judgment of what those proofs are worth.
Organization trust grows over time
Organizations that consistently issue well-scoped passports, maintain clean audit histories, and submit behavioral attestations accumulate issuer reputation. Higher-reputation organizations receive streamlined issuance and expanded fleet limits. Organizations with incident history face enhanced review.
Built open. Designed to federate.
Veris Protocol is centralized by design today — which is what makes it fast, reliable, and enterprise-deployable right now. The long-term vision is a federated verification network where multiple independent Veris nodes can cross-verify agent credentials without any single point of control.
- Single hosted verification service
- Fast, SLA-backed, enterprise-ready
- AI Passport credentials expressed as W3C DIDs for cross-platform portability
- OIDC compatibility
- Independent Veris nodes can cross-verify without central authority
- Multi-cloud
- IEEE / W3C working group candidacy
- Open standard adoption
Veris Protocol is published as an open RFC. We are committed to avoiding a new identity silo. Our interoperability roadmap is a protocol commitment, not a marketing promise — it is tracked publicly in the GitHub repository.
Veris does not prevent a bad actor from registering. It ensures that every registered agent is cryptographically attributable, continuously scored, and immediately revocable — making accountability unavoidable.
Veris in the wild
See how the protocol works inside real AI agent deployments.
A financial services firm deploys an AI research agent to analyze market data and generate investment reports. The agent needs access to sensitive financial data feeds but must be prevented from executing trades or modifying records. The compliance team requires a complete, queryable audit trail for every data access event.
1// Register the research agent once at deployment2const passport = await aegis.passports.issue({3 agentName: 'ResearchAgent-v3',4 modelFamily: 'claude-sonnet',5 deploymentEnv: 'production',6 permittedActions: [7 'read:market_data',8 'read:company_filings',9 'write:research_reports',10 ],11 deniedActions: [12 'write:transactions',13 'write:orders',14 'delete:records',15 ],16 requiresHumanApproval: ['read:client_portfolios'],17 jurisdiction: 'US',18});19// passport.passportId: ap_01HZ7K3N8P2Q4R5S6T7U8V9W0X20// passport.trustTier: TRUSTED (score: 850)1// Verify before the agent accesses the market data feed2const decision = await aegis.verify({3 passportId: passport.passportId,4 requestedCapabilities: ['read:market_data'],5 interactionType: 'data_access',6 riskLevel: 'HIGH',7 relyingParty: 'bloomberg-data-feed',8});910// decision.decision: 'ALLOW'11// decision.trustScore: 89212// decision.latencyMs: 231// Pull the audit log for the quarterly compliance report2const auditLog = await aegis.audit.getLog(passport.passportId);34// Returns tamper-evident, hash-chained records of every5// data access event — ready for SEC or FINRA review.Result: The compliance team can produce a complete, cryptographically verifiable record of every data access event in minutes — not days. The agent's scope is enforced at the credential level, not reconstructed from application logs after the fact.
Built for autonomous systems
Production-grade primitives mapped to NIST AI RMF and the EU AI Act.
AI Passport
Cryptographically signed agent identity with capability scope.
Trust Score
Real-time 0–1000 composite score across 5 signal families.
Trust Handshake
3-phase mutual verification before sensitive interactions.
Instant Revocation
<60s propagation across all verification nodes.
Audit Trail
Tamper-evident, hash-chained log of every agent action.
Compliance Ready
NIST AI RMF, EU AI Act, SOC 2 Type II mappings built in.