How it works
Three questions every service must ask.
Authentication: who is this agent? Both users and agents carry decentralized identifiers (DIDs): globally unique, cryptographically bound to a public key, with no central registration authority. The agent signs every request and the service verifies the signature against the agent's DID document, so impersonation isn't a question of trusting a platform; it's a question of math.
Delegation: who authorized this agent to act? The user grants the agent a verifiable credential (a delegation signed by the user's own wallet key) that defines the permitted scope and expiry. The agent presents the credential when it acts, and the service can independently verify the chain back to the responsible human across organizational boundaries, resolving trust against the DID method rather than a single central registry.
Authorization: what is the agent actually allowed to do right now? The service checks the credential's claims against the requested action. Where regulation permits, zero-knowledge proofs let the agent prove an action is within scope without revealing the policy itself, keeping commercially sensitive constraints like budget limits or allowed actions private from the verifier.
Further reading