Agent Identity

Distinguish the agent, the human or service it represents, and the credentials it uses. They are not the same thing.

Priority: P0
Phase: MVP
Primary owner: IAM / Platform
Status: Blueprint

Objective

Agents need first-class identity rather than borrowed human credentials. NIST’s 2026 concept work on software and AI-agent identity highlights identification, authorization, auditing and non-repudiation as emerging needs. Agent Trust Cloud should represent agent identity, the party it acts for, and the scope of delegated authority.

What to build

Create immutable Agent IDs and bind them to cryptographic/workload identity when possible.
Represent “acting on behalf of” relationships for humans, applications and other agents.
Support ephemeral session identities for individual agent runs or tasks.
Track identity assurance level based on registration source, attestation and credential strength.

Implementation decisions

Integrate with Entra, Okta and cloud workload identity rather than inventing a parallel human directory.
Avoid shared API keys as identity; if unavoidable, wrap them in a brokered identity context.
Include owner, purpose, environment and deployment metadata in authorization context.
Design identity federation so a customer can use its own certificate/workload identity infrastructure.

Definition of done

Every protected action maps to a specific agent identity and principal chain.
A compromised/retired agent can be revoked without disabling unrelated agents.
Identity context survives cross-service hops through signed claims or trusted session tokens.

Success metrics

Actions with verified agent identity
Shared-credential reduction
Identity assurance distribution
Revocation propagation time

IDENTITY RULE

Distinguish the agent, the human or service it represents, and the credentials it uses. They are not the same thing.

This control is specified on page 16 of the blueprint.