Barely. Agents authenticate through FleetWarrant instead of holding raw keys, and their traffic flows through a lightweight gateway — an SDK call or a sidecar proxy. The systems on the other side need no changes at all.

You plug FleetWarrant in and it watches. You see every agent running in your business, what it can reach, and what it did — without blocking anything. Most teams are surprised by the answer. Enforcement is a switch you flip when you're ready.

That's the point. Agents from OpenAI, Anthropic, Microsoft, Salesforce and your own team all sit behind the same identity, policy and audit layer. No model vendor can credibly police the others, so FleetWarrant stays neutral.

Credentials issued by FleetWarrant are short-lived and revocable at the gateway, so a revoke takes effect on the next call — typically under a second, and the audit log records exactly what was stopped.

Australian and EU regions at launch, with on-prem deployment and hardware attestation on the enterprise tier. Prompts and payloads can be redacted at the gateway before they ever reach our storage.

If you have one agent in production with real credentials, it isn't. Design partners get the audit-only wedge free, direct access to the team, and a say in the policy packs we ship first.