Model-agnostic agent identity
Works with any model or framework. The agent is a subject, not a vendor integration.
AI-Agent Trust
Identity, policy and scoring for autonomous agents.
AI-Agent Trust gives every autonomous agent a verifiable identity, a policy envelope describing what it may do, and a continuously updated trust score based on its behaviour. Agent actions are proof-backed, so an agent's work can be audited and, where a policy is breached, automatically constrained.
Built for: Teams putting AI agents in front of customers, money or production infrastructure.
Capabilities
Policy envelopes
Declare permitted actions, value limits and escalation rules. Breaches are recorded as evidence.
Behavioural trust scoring
Scores move with observed behaviour, so drift and anomalies surface before they cause loss.
Proof-backed actions
Each consequential action mints a proof, giving you a defensible record of agent decisions.
What you get out of it
- Let agents act autonomously with an auditable boundary
- Detect misbehaving or compromised agents automatically
- Show regulators and customers how agent decisions were governed
Try it right here
This runs the real logic on our servers — no account, no canned responses. Demo runs are not stored and use a published demo signing key; your tenant uses rotating ML-KEM-1024 material.
Authorise an agent action
Every autonomous action is scored against its mandate before it runs. Raise the amount or pick a sensitive action and watch the decision move from allow to step-up to deny.
How it fits the platform
AI-Agent Trust reuses Trust Kernel identities and Universal Proof artefacts, adding policy limits and behavioural scoring so autonomous actors operate inside a boundary you can audit.
See the four-step platform flow →Pricing
AI-Agent Trust is included from the Growth plan at $199 / month, with every higher plan including it too. Verification is never charged per check — plans differ on volume, retention, autonomy and governance.