Your AI spend is invisible.
Fix that in 60 seconds.
See where every dollar goes across ChatGPT, Claude, Copilot and more — and stop the waste. Starts right here in your browser: no signup, no install, and your usage data stays on your device.
One click shows a full real-shaped result. No terminal, no folder, no account.
I run agents / code with AI
Claude Code, Codex, OpenClaw, Hermes — catch runaway spend before it bites.
🛡️ Guard my agents →
I'm on a paid plan
ChatGPT Plus, Claude Max — hitting the usage bar? See what ate it and make the plan last.
💡 Beat the usage limit →
I buy AI for a team
Enterprise, government or consultancy — attribute every dollar to the team that spent it.
🏢 Show me the org view →
5.5B
tokens found on one machine
$4,245
of AI usage value, made visible
62×
a real tester's $40 plans returned, at API prices
60s
zero keys, nothing uploaded
Real result from the first machine ever scanned — costs at list prices.
Three ways in — all under a minute
1 · Scan
Pick your ~/.claude or ~/.codex folder in the browser. Transcripts parse locally — per project, per model, per day.
2 · Sync
Paste a read-only reporting key (OpenRouter covers Grok, DeepSeek, Qwen and more) and pull 30 days of provider-audited spend.
3 · Import
Drop a CSV from any provider console. Flexible headers, costs estimated when missing.
✓ Claude / Claude Code
local scan — zero keys
✓ ChatGPT / Codex
local scan — zero keys
✓ GitHub Copilot
seat import — idle seats in dollars
✓ OpenRouter
one key → Grok, DeepSeek, Qwen, Llama…
✓ OpenAI API
admin key sync (usage + Cost API)
✓ Anthropic API
admin key sync
✓ Any provider
CSV import
○ Gemini CLI
coming
○ Cursor
coming
Then the results do the talking
Token Waste Index
Cache misses, verbosity, frontier-model overuse, runaway agent spikes — each priced in $/month, claims capped, assumptions visible.
Attribution
Which project is burning the money. Spend by project, model, provider, and day — the answer finance keeps asking for.
Best model per task
A live quality-per-dollar index across 37 models and 11 providers, with a public leaderboard.
Built to be trusted
Zero float
We never hold your keys, credits, or money. Providers bill you directly; we bill for intelligence.
Local-first
Scans parse in your browser. Your usage data lives on your device — export or clear anytime.
Honest numbers
Independent counting, no pay-to-rank, savings claims capped — enforced by unit tests.