The simplest OpenCode usage tracker is UsageAtlas, a free and open-source desktop app that reads the database OpenCode already keeps on your computer and turns it into tokens, requests, models, sessions and an estimated cost. OpenCode's own opencode stats command is the quick terminal check, and UsageAtlas adds the daily view, the history and the other coding tools you use.
Why OpenCode needs a tracker at all #
OpenCode is a frontend for models from many providers. One week you might run a Claude model through one key, a GPT model through another and a local model in between. Each provider bills you separately, so no single page tells you how much OpenCode work you did or which model it went to.
The good news is that OpenCode records this on your machine. Every session leaves token counts and model names in a local database. A tracker's job is to read that record and show it in a form you can scan in a few seconds.
What UsageAtlas shows for OpenCode #
UsageAtlas opens that local database and fills its standard views with your OpenCode numbers:
- Today : tokens, requests and models for the current day.
- History : the same figures day by day, so you can spot a heavy week.
- Insights : model share, plus time-of-day and weekday patterns.
- Cost : an estimate built from tokens and model pricing, labeled as an estimate. When OpenCode reports a cost, that figure is used instead.
A per-tool filter narrows everything to OpenCode. Turn the filter off and OpenCode sits next to Claude Code, Codex, Cursor, GitHub Copilot and the other supported tools in one total.
opencode stats or a dashboard #
| Question | opencode stats |
UsageAtlas |
|---|---|---|
| Tokens and cost for recent days | Yes | Yes |
| Breakdown by model | With --models |
Yes |
| Filter by project | With --project |
Filter by tool |
| Day-by-day history you can browse | No | Yes |
| Other coding tools in the same view | No | Yes |
| Runs without a terminal open | No | Yes |
OpenCode's CLI docs describe opencode stats as token usage and cost statistics for your sessions, with --days to limit the range. If one number once a week is all you want, that command is enough and you need nothing else.
Set it up in four steps #
- Use OpenCode on this computer at least once, so the local database exists.
- Download UsageAtlas for Windows, macOS or Linux from the UsageAtlas home page .
- Open the app. It finds the OpenCode data on the same machine; there is no separate OpenCode login.
- Check Today, then come back to History after a few days of work. Use the tool filter if other coding tools are connected.
No account is needed for local tracking. A free Google or GitHub sign-in is optional and only saves usage summaries so you can restore them on another computer. Prompts and responses are not read, and project and session names stay out of cloud saves.
What the numbers are and are not #
UsageAtlas is a view of your local usage, not a bill. Its cost column is an estimate, and it can differ from your invoices because of cache pricing, provider differences or late billing. When the amount has to be exact, use each provider's own invoice. OpenCode's local history also does not carry provider quotas, so treat the tracker as a record of what you used rather than a countdown to a limit.
The short version #
If you live in OpenCode and want to know where your tokens go, install UsageAtlas on the same machine and let it read what OpenCode already stores. Keep opencode stats for a quick look from the terminal, and keep your provider invoices for the final money figure.