01
What ccusage gets right
ccusage is a genuinely good free CLI: run it with npx and it reads the usage records coding agents keep on your machine, then prints daily, monthly, and per-session tables with cost estimates. It started with Claude Code and now covers many agent CLIs. For the question "what did last month cost me", it is a fine answer, and being open source under MIT, it costs nothing to try.
02
A report is not a monitor
The difference is pull versus push. A CLI report shows usage when you remember to run it, which is usually after a limit already landed. AgentPeek keeps live gauges in the Mac notch and menu bar: 5-hour and 7-day windows for Claude Code and Codex with reset countdowns, Cursor's monthly window, and token totals for the local-log agents.
- Budget alerts fire once at your threshold, before a limit interrupts a refactor.
- Limit windows come from what each agent itself reports, so no quota is ever invented.
- A by-day chart and input, output, and cache split cover the history questions in the same panel.
03
Usage is one page of it
AgentPeek treats usage as one surface among many. The same app shows every live session with its state, transcript, diffs, and todos, surfaces permission prompts you can answer with a keystroke, sends the next prompt into a resumable session, sorts everything on the Agent Board, and lists your local dev servers. A usage CLI is one tab of that job.
04
Run both if you like
This is not either-or. ccusage and AgentPeek read the same local records, so they never conflict: keep ccusage for scripted exports and terminal deep dives, and let AgentPeek handle the always-on part. If you are weighing the cost, the build vs buy page has the honest math on what $15 buys.