As AI agents become critical actors in digital ecosystems, we need standardized ways to represent and verify their identities. To address this, I’ve proposed OpenID Connect for Agents (OIDC-A) 1.0 — an extension of OpenID Connect Core 1.0 specifically for agent-driven interactions.
OIDC-A defines:
1. Standardized agent identity claims (model, provider, version, instance ID)
2. Delegation chains for authority tracking across agents
3. Cryptographic attestations to validate agent integrity
4. Capability discovery mechanisms to evaluate trust and permissions
The goal is to enable secure, verifiable, and interoperable identity for agents operating independently on behalf of users or systems.
Would love feedback from the security, identity, and AI communities.
As AI agents become critical actors in digital ecosystems, we need standardized ways to represent and verify their identities. To address this, I’ve proposed OpenID Connect for Agents (OIDC-A) 1.0 — an extension of OpenID Connect Core 1.0 specifically for agent-driven interactions.
OIDC-A defines:
1. Standardized agent identity claims (model, provider, version, instance ID) 2. Delegation chains for authority tracking across agents 3. Cryptographic attestations to validate agent integrity 4. Capability discovery mechanisms to evaluate trust and permissions
The goal is to enable secure, verifiable, and interoperable identity for agents operating independently on behalf of users or systems.
Would love feedback from the security, identity, and AI communities.
Proposal here: https://subramanya.ai/2025/04/28/oidc-a-proposal/