Fleet AI Gateway
Optional pooled model capacity without making a proxy mandatory for AgentOS.
The Fleet AI Gateway is optional pooled capacity for reviewed Codex subscription accounts with an explicit OpenAI API-key fallback. Direct per-Agent authentication remains a complete AgentOS topology.
Use the owning procedure before selecting, installing, authenticating, recovering, rotating or retiring the gateway. It keeps server-owned provider OAuth and quota-aware routing inside a deliberate service boundary. Client Agents receive only the scoped connection material intended for them; they never receive a provider credential or Fleet-shared inference token.
Clients may live in agentos or a Second-Mate domain namespace. They call
agentgateway-openai.agentos.svc.cluster.local and authenticate with their kubelet-rotated,
audience-bound ServiceAccount token. Agentgateway asks the TokenReview/OpenFGA authorizer for a
closed grant lasting at most 15 seconds. Only Agentgateway can reach the AI Gateway through its
NetworkPolicy; provider OAuth, the operator Secret and the Gateway vault remain in the core
namespace. This lets a Second Mate manage Crewmates in its namespace without gaining another
domain’s identity or provider credentials, while unrelated Internet traffic stays direct.
What it does not own
The gateway does not create Agent identity, Assignments, human authority or provider-neutral model semantics. It is not a universal LLM abstraction and does not make one account’s quota an organizational permission.
Inspect accounts, health, routing state and observed provider failures through the service’s
current interfaces. A 401, 403 and 429 have different causes; recovery does not silently swap
identities, bypass OpenFGA or introduce an API-key fallback that was never approved.
Because the feature is optional, every core Docs path remains understandable without it.
Last updated on