Monitoring Claude Code usage: three ways compared
There are three common ways to monitor Claude Code usage. First decide whether you want a report afterwards or a stop at the moment an agent gets stuck.
Data as of 2026-10-31
The three side by side
| /usage | ccusage | Budgethelm | |
|---|---|---|---|
| Which tools | Claude Code | Many coding agent CLIs (per its README) | Claude Code, Codex, Gemini CLI, Copilot CLI, Cline and more |
| Which computers | Local records on this one | Local data on this computer | Every computer you install it on, in one place |
| Where data goes | Stays local | Stays local | Only time, tool, model, token counts and stop records are uploaded |
| When an agent gets stuck | You look it up afterwards | You read the report afterwards | It is stopped right then, with a desktop notification |
/usage follows the Claude Code docs; ccusage follows its README.
What Budgethelm shows
- Spend: tokens, cost and error rate by project, agent and model, with a monthly budget warning. How cost monitoring works
- Waste: repeated context, no prompt caching, overly long output, expensive models on simple tasks. Claude Code savings
- Stop records: when and why an agent was stopped, and how many calls that saved.
How to choose
- One computer, Claude Code only: start with
/usage. - Want open source and local reports: read the ccusage and Budgethelm comparison.
- Several tools, several computers, or unattended
claude -pruns: add Budgethelm.
FAQ
- Does Claude Code have built-in usage monitoring?
- Yes. /usage shows local records on this computer (per the official docs, excluding other devices).
- How far back does the Budgethelm dashboard go?
- The dashboard defaults to the last 14 days, and cost analysis accepts any date range.
- Does it cost anything?
- It is free now: sign up, no credit card. Paid plans are planned; see Plans on the home page.