Usage stats
Press ⌘U (Ctrl+Shift+U on Windows) to open the usage panel and see which model and which project your tokens went to.
Where the data comes from
Belfry reads and aggregates the local Codex / Claude session logs directly, without making any network requests. The data never leaves your machine.
Four token categories
Belfry normalizes tokens into four categories, so that different CLIs computing things differently doesn't muddy the numbers:
| Category | Meaning |
|---|---|
input | Input tokens (excluding cache hits) |
cachedInput | The portion that hit the context cache |
cacheWrite | Writes to the context cache |
output | Output tokens |
input and cachedInput are separated because cache hits are usually billed at a much lower rate — merging them would overstate your real spend.
Two breakdown dimensions
- By model — see which model costs the most
- By project — see which project costs the most
Time windows: last 7 days / last 30 days / all time.
Quotas and plans
If the logs carry quota window and plan information, the panel shows it too. Note: currently only Codex's logs carry these fields — Claude's do not, so you won't see quota information for Claude sessions.
The usage panel's numbers come from the local logs written by the Agents themselves, so treat them as a reference, not a bill. Always go by your provider's console.