Reading Codex usage on your Mac

By MacStick · Updated 2026-11-15 · 2 min read

AI workflow illustration; example interface

Know which source you are reading

MacStick reads local Codex CLI session records. The view describes the records available on this Mac, not every project or session associated with an account. The browser demo is a visual example and never reads your Codex files.

Cumulative totals need careful handling

Some session records contain a running total. Adding every running total together would count the same work repeatedly. The implementation computes changes between cumulative totals when folding a session into the usage ledger. This is why a raw line count is not a useful proxy for token use.

Usage windows are separate from tokens

When a Codex record contains rate-limit information, the app can display what that record reports. That information may be absent or stale. A remaining allowance should not be inferred from a token count, and an empty field should not be described as unlimited access. Check the provider’s own interface when capacity matters.

Use estimates with context

For useful comparisons, look at task size, model, available session history and time period together. A long investigation and a short edit are different workloads. Recorded Codex limit windows are a separate signal from token counts; neither is an invoice or a promise about future plan limits.

Choose reporting or interaction deliberately

If you need to approve tool calls or review agent plans from the notch, seek a product that explicitly supports those actions. MacStick currently concentrates on local usage visibility alongside everyday Mac utilities. See the comparison directory for agent-focused alternatives and verify their current compatibility before installing.

See the idea in action.

The browser preview uses sample data and does not access your Mac.

Try MacStick →