--- title: "Track cost" description: "The running cost chip, per-run usage attribution, credit holds, and the BYOK badge" canonical_url: "https://hoplite.sh/docs/threads/cost" markdown_url: "https://hoplite.sh/docs/threads/cost.md" --- # Track cost URL: /docs/threads/cost LLM index: /llms.txt Description: The running cost chip, per-run usage attribution, credit holds, and the BYOK badge Related: /docs/workspace/billing, /docs/agent/models, /docs/threads/manage # Track cost Each thread shows a running cost chip, so you always know what a piece of work has spent. The detail behind it lives on the usage page and in your workspace's [billing](/docs/workspace/billing). ## The cost chip The chip on a thread tracks its running cost as the agent works. Every model call is attributed to the run that spent it, so the number reflects the thread's actual usage. ## Per-run attribution The **Usage** page charts spend per day and breaks usage down by model and run, so you can see exactly where credits go. Every run's token usage is attributed there. ## How credits settle Runs reserve credits while in flight and settle to actual usage when they finish. On the billing page these reservations show as **Held** — released back, minus actual usage, when the run completes. See [Billing](/docs/workspace/billing). ## The BYOK badge In the cost breakdown, a **BYOK** badge beside LLM inference means every run with a durable model-resolution record used the workspace's provider credentials. Mixed threads and historical usage without a resolution record stay unbadged, so the UI does not overstate who supplied credentials. Runs on Codex models bill your connected ChatGPT plan instead of workspace credits. Their usage still shows in your usage history — with zero credit cost. See [Models](/docs/agent/models). ## Sitemap See the full [sitemap](/docs/sitemap.md) for all pages. Well-known sitemap: [/docs/.well-known/sitemap.md](/docs/.well-known/sitemap.md).