Best fit
- You want AI apps to keep one stable local endpoint.
- You need provider switching, request logs, cost visibility, or failover.
- You want client config changes to be reversible.
Not ideal
- You use one client, one provider, and the setup rarely changes.
- You do not need logs, failover, cost tracking, or config rollback.
The problem
When a client talks directly to a provider, failed requests and cost spikes are hard to inspect. AgentGate makes the request path visible before it leaves your machine.
What AgentGate records
You can inspect route profile, upstream provider, model, status, latency, token usage, estimated cost, and failover behavior when available.
How to use it
Apply the Codex config, run a request, then open the Logs page. Filter by client or provider to understand which route handled the request.