Make agent activity visible

Identify agent traffic with explicit headers and log what each agent did, when, and why.

Why

Without differentiating agent traffic from human traffic, services cannot debug failures, audit what agents did, or improve their AX based on real usage. The canonical AX framework (agentexperience.ax — Differentiate Agent Interaction) states this explicitly: "digital services will be harder to debug, lack context for usage audits, and will miss key insights on how agents use their service." Speakeasy identifies Traceability as one of four foundational AX properties for APIs. The canonical agent-facing principle (Transparent Identity) requires agents to identify themselves via User-Agent — this principle is the service-side mirror of that requirement.

Do

Require X-Agent-Name and audit-log agent actions.

Don't

Let unidentified agents act invisibly.

Artifact

X-Agent-Request: true
X-Agent-Name: acme-support-bot
X-Agent-Operator: org_acme
User-Agent: acme-support-bot/1.4 (+https://acme.com/agents)

{ "ts": "2026-06-04T12:00:00Z", "agent": "acme-support-bot", "operator": "org_acme",
  "action": "create_refund", "resource": "ord_123", "result": "ok", "reason": "duplicate charge" }