Track usage and savings

Because every tool routes through one gateway, the dashboard can show you something no single provider can: every request across every account in one view, in real time, and what that usage would have cost at metered API rates. This guide covers the two things people watch most: live request detail, and savings.

Every request, in one place

Open the dashboard and you get a single feed of requests across all your connected providers, updating live as your tools work. For each request you can see the model it resolved to, the provider it ran on, token counts, whether it fell back to a backup provider, and its cost. It is the whole point of having one endpoint: your Claude Code, Codex, and Raycast usage stops being three separate bills you never reconcile and becomes one timeline you can actually read.

This is useful in the moment, where you can watch an agent run and see exactly which model answered each turn, and after the fact, for spotting which tool or model is eating your budget.

The savings index

OpenLLM adds no markup, so the cost it shows is the real provider cost. When you run on a subscription, that cost is effectively zero per request, since you have already paid the flat monthly rate. The dashboard makes that concrete by comparing two numbers:

• What you actually paid: for subscription usage, this rides your flat rate.
• API-equivalent cost: what the exact same requests would have cost billed at the provider's metered API price.

The gap between them is your savings: the double-billing you would have paid if you drove these agents on metered API keys instead of the subscriptions you already own. The subscription pricing index tracks this per provider, including how much of your plan's window you have used and how much headroom is left, so you can see both what you saved and how hard you are leaning on each subscription.

Reading it from the API

The same figures are available on the gateway if you want them in a script. GET /stats returns your totals and breakdowns:

curl https://openllm.sh/stats \
  -H "Authorization: Bearer sk-llm-YOUR_KEY"

The response includes total_requests, total_cost_usd (what you paid), total_api_equivalent_cost_usd (the metered-API comparison), a per-model breakdown under by_model, a per-day series under daily, and the subscription index under inferred_subscription_usage. Point a small dashboard or a cron report at it and you have your own savings tracker.