Anchorage API reference
    Preparing search index...

    Interface AgentAuditContext

    Trusted scalar correlation attached to agent authorization and policy events. Hosts must derive these values instead of accepting them from prompts, tool inputs, or other untrusted payloads.

    interface AgentAuditContext {
        agentId: string;
        delegatedBy?: string;
        entryPath: string;
        principalId?: string;
        principalKind?: string;
        purpose?: string;
        resourceId?: string;
        runId?: string;
        tenantId?: string;
        threadId?: string;
    }
    Index

    Properties

    agentId delegatedBy? entryPath principalId? principalKind? purpose? resourceId? runId? tenantId? threadId?

    Properties

    agentId: string

    Guarded agent identifier.

    delegatedBy?: string

    The principal that delegated this execution, for agent-to-agent work.

    entryPath: string

    Trusted invocation path, such as an HTTP start or approval resume.

    principalId?: string

    Stable principal identifier. Distinct from the audit event's actor.id, which is the identity the gate evaluated: on an approval resume the actor is the restored original principal while the decision belongs to a human reviewer, and a correlated trail needs to name both.

    principalKind?: string

    Which kind of principal is executing: human, service, agent, or system.

    purpose?: string

    Why an automated principal exists, as declared by the host.

    resourceId?: string

    Server-derived memory resource identifier.

    runId?: string

    Server-minted run identifier.

    tenantId?: string

    Tenant identifier asserted by the authenticated host.

    threadId?: string

    Server-minted or tenant-owned thread identifier.

    Settings

    Member Visibility

    On This Page

    Properties