AWARE
The security control plane for autonomous AI agents. Establish agent identity, enforce policy on every tool call, revoke authority when it must end, and keep evidence a third party can check.
Five capabilities
Identity
A non-human identity per agent, with a credential stored only as a salted, peppered hash and compared in constant time.
Governance
Policy enforcement on every tool call, deny by default, evaluated against tool risk and data classification tier.
Delegation
The intended model bounds a child’s authority by its parent’s. The delegation graph that would carry this is roadmap.
Revocation
Revoking an agent invalidates its credential immediately and records the decision. Cascade to derived agents is roadmap.
Assurance
Auditability and evidence are implemented. Independent execution verification is roadmap, and labelled so.
Compliance
Seven frameworks mapped, 184 of 243 CSA AICM v1 controls, MITRE ATLAS technique mapping. A mapping is not a certification.