Kibble Blog

ccusage for a team: what it covers and where it stops

By Kibble · 5 September 2026 · 5 min read

ccusage can answer what a machine's coding agents used. For a team, you also need a reliable way to collect reports, map them to people, control access and explain gaps. You can build that around local tools, use a community team extension, or use a shared collector such as Kibble.

Does ccusage support Claude Code and Codex?

Yes. The current ccusage guide covers Claude Code, Codex and other coding CLIs. It provides local usage reporting from the data those tools keep. Calling it Claude-only would miss its current scope. Check its documentation for the supported commands and data locations in the version you use. As of September 2026 it also has an experimental GitHub Copilot CLI source that reads the same shutdown counters Kibble's Copilot collector reads, on one machine and without Copilot Chat in VS Code.

Local counts priced at API rates are estimates. On a subscription, they describe equivalent API usage rather than a second bill. That distinction still matters after several local reports are added together.

What are the choices for a team?

Community options exist. For example, the ccusage-team package listing describes live team monitoring. Its linked source repository was unavailable when checked, so verify accessible source, maintenance and what the chosen version sends before adopting it. A package name alone establishes neither privacy nor security. This comparison was checked on 5 September 2026.

When is a local tool enough?

Use the local report when each person needs their own answer, or when your team can maintain a small aggregation workflow. Agree on the date window and time zone, preserve the model and token categories, and distinguish missing reports from genuine zero usage. Do not collect raw transcripts just to add token counts: those files can contain prompts and code.

CI can also produce supported logs. Decide how a runner reports before it disappears and how it maps to an identity. Neither installing a local tool nor adding a web dashboard automatically guarantees that ephemeral runners are covered.

What does Kibble add?

Kibble (usekibble.com) is a Claude Code and Codex usage dashboard for teams. Its open-source collector derives daily summaries from local logs and pushes them to a shared dashboard. Owners see organization totals and member breakdowns; managers see assigned teams; members see themselves. Per-machine details remain visible only to the machine's owner.

Kibble also merges covered Claude Console vendor rows with local estimates, giving vendor estimates precedence over matching local rows at the reporting grain. Repository figures remain a parallel local estimate. Kibble does not turn ccusage output into invoices or reconcile Codex usage to an OpenAI bill. Read estimated versus billed before comparing totals.

How do you try team reporting?

1. Install Kibble with Node.js 20+, then run kibble login on a machine with supported logs.
2. Run kibble push for the first counts and kibble schedule status to check background collection.
3. Compare the same day's model and token totals with your local report. Investigate any difference in coverage or rates.
4. For a team rollout, use Pro invitations and assign reporting teams. Add the organization's Claude Console key if vendor coverage is needed.

Free covers one person and two machines with 30 days of history. Pro covers up to 30 people for $100 monthly or $1,000 yearly, before tax, with a 30-day first-subscription trial and a card required. See current pricing for additional seats.

For each option, ask the same question: can you explain which people and machines contributed to the total, what is missing, and whether the number is estimated or billed? Choose the simplest setup that answers it.