For teams shipping with Claude Code
Know what your team actually spends on AI coding agents.
AgentSpend turns everyone's local Claude Code usage logs into one team dashboard — with budget alerts, before the bill surprises you.
Free for solo use. No credit card. We never read your code or prompts — see how.
The problem
“I have no idea which of my engineers are heavy Claude Code users until the bill lands at the end of the month.”
“Everyone runs their own local CLI usage tool, but nobody on the team can see the combined picture.”
“I want a heads-up before we go over budget, not a console bill I find out about after.”
How it works
Install the CLI
One command on each teammate's machine. It reads your existing local Claude Code logs — nothing new to configure.
Connect your team
Everyone points at the same team key. Usage starts flowing into one shared dashboard within minutes.
See spend, set budgets
Daily trends, per-member and per-project breakdowns, and email alerts before you blow through budget.
One dashboard for the whole team
Stop asking teammates to paste their local usage numbers into a spreadsheet. See everyone's spend in one place, updated automatically.
Budget alerts
Set a monthly ceiling per team or per project. Get an email before you go over — not an invoice after.
History that survives log rotation
Local logs get cleaned up or a laptop gets wiped. Your usage history lives in the dashboard, not just on disk.
Your code never leaves the machine
The CLI reads only token counts, model name, project label, and timestamps from your logs. Prompt and file contents are never read or transmitted.
How it compares
Free local tools like ccusage are great for one person. Other team dashboards exist too — here's where AgentSpend lands.
| Free CLI (ccusage) | Other paid dashboards | AgentSpend | |
|---|---|---|---|
| How it collects | Reads local logs | Sits in the request path to route or compress traffic | Reads local logs — nothing sits between you and the API |
| Collector source | Open source | Closed source | Open source — run --dry-run to see the exact payload |
| Team dashboard | — | ✓ | ✓ |
| Budget alerts | — | ✓ | ✓ by email |
| History survives log cleanup | — | ✓ | ✓ |
| Price | Free | From ~$15/mo | $19/mo, up to 10 people |
What we actually collect
Some team dashboards work by putting themselves between you and the model — you run their command instead of claude, and your file reads and build output pass through their servers. AgentSpend never does that. It reads the log files Claude Code already wrote to your disk, after the fact.
This is the entire extraction step — every field that ever leaves your machine:
events.push({
occurred_at: entry.timestamp,
project_label: path.basename(entry.cwd), // folder name only
model: entry.message.model,
input_tokens: usage.input_tokens,
output_tokens: usage.output_tokens,
cache_read_tokens: usage.cache_read_input_tokens,
cache_creation_tokens: usage.cache_creation_input_tokens,
});No prompt text, no file contents, no diffs — there is no line in the collector that reads them. Run npx github:THEMANJH/agentspend-upload --dry-run to print the exact payload before sending anything, or read all ~200 lines of it. Zero dependencies, MIT licensed.
Pricing
Team
$19/mo
- Up to 10 members
- Unlimited history
- Budget alerts by email
- Per-member & per-project breakdown
No setup fees, cancel anytime. Team unlocks the moment payment goes through.