Everything your engineers say to an agent never leaves their laptop.
Made-up sessions on an example team. Kibble never reads what an agent is told.
The command center for AI-native companies.
Every token, every seat, in one bowl
- this week
- $122.54
- against last
- ×2.2 ($56.40 last week)
- biggest mover
- Mai, web, ×4.0
- engineers
- 5
- sessions
- 16
- tokens
- 143.4M
- models
- claude-opus-5, claude-sonnet-5
Per person, team, model and repo. From counts, never content.
What you have.
One invoice per vendor, one number per month, after the month is over. It counts seats, not usage. It cannot say which engineer, which repo, which model, or what changed.
So the question lands on you before the numbers do: why did it double?
What the counts give you.
- Who burned what: spend per engineer, team, model and repo, this week against last, from every laptop, CI runner and cloud box.
- Billed truth from your Claude Console, merged with the local estimate and never summed with it.
- A small open-source CLI reads the logs the agents already write and pushes a per-day summary every hour.
- Visibility and alerts, never surveillance. Each engineer sees their own machines; you see totals.
This week, one line per session, by its opening prompt. Example content
That was the last session. None of it left the laptop. Only the counts crossed.
Who burned what.
| engineer | repo | sessions | tokens | spend |
|---|---|---|---|---|
| team | 0 | 0 | $0.00 |
- by repo
- nothing yet
- by model
- nothing yet
The sessions are made up. The arithmetic is real: this page adds them up as they cross, priced at the list rates the CLI uses (Opus 5 at $5 in, $25 out per million; Sonnet 5 at $2 and $10; cache reads at a tenth).
Prompts. Files. Paths. Diffs. Hostnames. Still here.
Nothing else leaves.
The server's ingest schema is strict: a field it does not expect is a rejected request, not a stored one. Every line the collector sends is in the open-source repo, and you can read it.
Content stays.
Why it doubled.
The answer is on the screen before anyone asks: who, in which repo, on which model, against last week.
| engineer | repo | last week | this week | change |
|---|---|---|---|---|
| team |
- Free
- One person, two machines, every agent, the last 30 days of history.
- Pro
- $200 a month or $1,800 a year for 50 people, then $5 per person a month. All history, teams, roles, invitations, billed truth from your Console. Before tax.
- Self-hosted
- On the roadmap. Kibble on your own infrastructure, every feature on, nothing metered, no key to enter.
npm install -g @usekibble/cli && kibble login # the first numbers land in about a minute
Questions.
The short answers, in one place. The long ones are in the collector's source.
- What does Kibble show?
- What every coding agent on your team spent this week: spend per engineer, team, model and repository, this week against last, with the biggest mover named. It covers every laptop, CI runner and cloud box that runs the collector, plus billed truth read from your Claude Console.
- Which coding agents does Kibble track?
- Claude Code and Codex from their local usage logs today, and your organization's Claude Console for billed usage. Anything the Anthropic Admin API bills is covered by the vendor sync.
- Does Kibble read prompts or code?
- No. The collector sends counts only: token totals, model names, opaque session ids and repository names. Never prompts, file contents, tool arguments, paths, hostnames or hardware ids. The server's ingest schema is strict, so a field it does not expect is a rejected request. The collector is open source and short enough to read.
- How is Kibble installed?
- Sign in with Google at app.usekibble.com, then on each engineer's machine run
npm install -g @usekibble/cli && kibble login. The CLI reads the logs the agents already write and pushes a per-day summary every hour. The first numbers land in about a minute. - How much does Kibble cost?
- Free covers one person, two machines, every agent, and the last 30 days of history. Pro is $200 a month or $1,800 a year for 50 people, then $5 per person a month, with all history, teams, roles, invitations and billed truth from your Console, before tax. Self-hosting is on the roadmap: every feature on, nothing metered, on your own infrastructure.
- Is the local spend number the same as the invoice?
- Local rows are estimates priced at the vendors' list rates. Billed rows come from your Claude Console and are the truth. Kibble merges them per day, billed wins, and the two are never summed.
- Can engineers see each other's usage?
- Each engineer sees their own machines. Owners see totals per person, team, model and repo. Managers see the teams an owner assigned them. Devices are visible to their owner only.
- Can Kibble be self-hosted?
- Not yet. Self-hosting is on the roadmap: Kibble on your own infrastructure, every feature on, nothing metered. The CLI collector is open source today, so what leaves each machine can already be read.