AgentOracle adds a check before an AI agent takes action. It looks at the claim the agent is relying on, compares it with sources and returns a verdict, such as act or do not act, with a confidence score.
Each result comes with a signed receipt that anyone can verify offline against published keys, rather than relying only on internal logs. It uses Ed25519 signatures and RFC 8785, and relates to an IETF draft. It is a paid service for teams that need provable agent decisions.
