When to use agent keys
By Aditya Perswal
When should you use agent keys? When you run 24/7 workloads out of tools like Cursor, Claude Code, Codex, or OpenCode.
Take a daily brief. Do you really care if the brief you read at 8:00 AM was created at 7:59 or at 7:55? It doesn't matter if it took five minutes to make or one. You got your brief. You would care if it was 50% cheaper. When you use an agent key it stores the deadline, so every request from an always-on workload tries the cheaper tier first without a code change.
A store for your logs
When you want to use FlexInference as your store for logs, an agent key is a great place for that too. Every request through the key is on one page. It shows each request with its model, token counts, cost, time to first token, and tier. We store traces only if you ask us to.
One bill for the team
When you want centralized billing for your whole team, agent keys do that too. Give each teammate and each agent its own key under one organization. Everything bills to one balance, and the logs say which key spent what. If you bring your own key (BYOK), you don't even pay the 10% top-up fee.
Closed and open models on one key
When you want to switch between closed and open source models, it's just a click. The key stays the same while the model name changes, so trying a model that costs a hundredth as much is a one-word change.