Tokenwise gives engineering teams per-developer, per-project Claude Code spend visibility — with Slack alerts before anyone burns through their budget.
Same problem, different context — Tokenwise handles all of them.
Your team spends $800-3,000/month on Claude API calls. Right now that's a single line item. Tokenwise breaks it down by engineer and feature area.
You're using Claude on client projects but charging flat rates. Tokenwise gives you timestamped per-project API costs to charge back accurately or defend your margins.
You're building on top of Claude API. You need per-user consumption data to set pricing tiers before your bill grows 4× from a handful of power users.
No agent to deploy. No code changes. One curl command from your Claude Code workflow.
Create an account and grab your per-team ingest endpoint. Takes 60 seconds — magic link, no password.
Add one line to your post-session workflow or CI pipeline. Claude Code session stats posted via HTTP.
curl -X POST https://tokenwise.fyi/ingest \
-H "Authorization: Bearer $TW_TOKEN" \
-d @session.json
Spend by developer, project, and day — visible to the whole team. Set alerts, export CSV, or view burn rate trends.
Every developer's token usage and dollar cost in one view. Know immediately when Alice's session costs 10× the team average.
Set thresholds per-user or per-org. Get a Slack message before anyone burns through their monthly allowance — not after.
Rolling spend rate + daily breakdown so you can forecast the month-end bill before it arrives. Per-project splits available.
Pull your team's full session history as CSV. Client billing, cost attribution, or finance team requests — covered.
You POST session data (token counts, model, duration, project tag) to your team's ingest endpoint. One curl command. Works from any CI pipeline, post-session hook, or editor extension. Your sessions are never sent automatically — you control what's uploaded.
Tokenwise accepts any session data you send. Claude Code is the primary use case, but if you're using Cursor, Windsurf, or direct API calls, you can ingest those too — cost is calculated from raw token counts at published Claude Sonnet 4.6 pricing.
No. Tokenwise is an independent tool. You still pay Anthropic directly for Claude usage — Tokenwise just makes that spend visible and accountable across your team.
Cost is calculated from raw token counts you provide. As long as you send accurate input/output token counts, the numbers are exact. Model pricing is updated when Anthropic announces changes.
Each team member gets their own ingest token and shows up as a separate row in the dashboard. Starter supports 5 seats; Growth supports 15. Everyone logs in via magic link — no passwords, no SSO setup.
Yes. Cancel from the billing portal — no questions asked. Your data stays accessible until the period ends.