How to track Claude Code costs per developer and team
Where per-person Claude Code spend shows up for each kind of account, why mixed accounts and cloud sessions fall through the gaps, and how to see one total per Member, Project and Device.
Claude Code is billed by the token, and per-developer spend varies a lot. Anthropic's own cost guide puts the enterprise average at around $13 per developer per active day and $150–250 per developer per month, with 90% of users under $30 per active day. The spread is the reason to measure: a handful of long Opus sessions can outweigh the rest of a team.
What each account type already gives you
Where the per-person numbers live depends on how each developer signs in to Claude Code:
| How a developer signs in | Where spend shows up | Per person? |
|---|---|---|
| Claude for Teams or Enterprise | Spend report in claude.ai org analytics, for usage credits past the seat allowance | Yes, inside that organization |
| Claude Console (API key) | The Console usage page and Claude Code dashboard, or the Claude Code Analytics API | Yes, inside that Console account |
| Bedrock, Google Cloud, Foundry | Your cloud billing console | No, unless you add OpenTelemetry or a gateway |
| Personal Pro or Max | claude.ai Settings > Usage and the /usage plan bars: account-wide usage, no per-session dollars | Only their own |
Each of these is accurate for what it covers. The gaps open when a team does not fit in one row.
Where per-person tracking breaks
- Mixed accounts. One developer is on a personal Max plan, another on the company's Console key, a contractor on their own. No single vendor dashboard sees all three.
- More than one machine.
/usageand local tools such as ccusage read the files on the machine they run on. A laptop, a dev server and a cloud session are three separate answers that nobody adds up. - Cloud sessions. A Claude Code cloud container is discarded after the session, and its usage files go with it. Nothing on anyone's laptop records it.
- Cost by repository. Vendor reports group by person and model. "What did the payments service cost us this month?" needs each request tied to the repository it ran in.
Track it with SessClone
SessClone puts a small plugin, the Collector, on each machine. It reads the usage Claude Code already writes, and reports token counts, models, timings and the git remote of each Turn to your Org. It never sends a prompt or a file's contents, unless a Member opts in to transcript archival. Because it reports from the machine, it does not matter which account a developer signs in with.
- Create an Org at sign up, or self-host it for free.
- Install the Collector on each machine: two commands in Claude Code and a restart. See Install the Collector.
- Add it to cloud environments too, so cloud sessions are counted. The same plugin runs from a cloud environment's setup script, and every container a Member uses collapses into one Device. This needs Claude's API credentials for cloud environments, which are offered on Pro and Max plans only for now.
- Invite your team. Each person issues their own API key under Keys, so their Turns land on them.
Costs then reads the same data four ways:
- Over time: a bar per day, opening that day's tokens by class and model, then the Sessions that ran.
- People: one row per Member, with their token table, a breakdown by model, and their Sessions.
- Projects: one row per repository, keyed on the normalised git remote, so spend lands on the code that caused it.
- Devices: one row per machine or cloud environment.
Roles decide who sees what. Owners and Admins see the whole Org, a Manager sees only the Members in their Scope, and a Member sees their own usage.
What the number means
Every cost in SessClone is an estimate: the Turn's tokens multiplied by the published per-token rate for its model at the time it ran. It is not your invoice. For subscription users it answers "what would this have cost on the API?", which is the figure to compare when deciding whether a seat, a Max plan or an API key is cheaper for a given person. A Turn on a model with no rate shows as unpriced rather than $0.
Claude Code cost per session shows the arithmetic, and cache tokens explained covers the token class that usually dominates the bill. To plan a monthly figure and set caps, see how to budget Claude Code for a team.
Is SessClone the right tool?
If you only need your own spend on one machine, npx ccusage@latest is
quicker. If everyone is in one Claude organization, Anthropic's analytics may
be enough. SessClone vs ccusage vs Anthropic's analytics
compares them side by side. Plans and prices for the hosted service are on the
pricing page; self-hosting is free at any size.
SessClone vs ccusage vs Anthropic's Claude Code analytics
Three ways to see what Claude Code costs, what each one covers, and when SessClone is the one you want.
Claude Code cost per session, and how it is calculated
How a Claude Code session turns into dollars — the five token classes, the per-model rates, the modifiers for fast mode, US-only inference and batch — with a worked example you can check by hand.