Usage per day
Accurate count, split by tool. Hover a day for details.
Monthly budget
Set a budget in Prices & budget.
Where tokens go
Fresh input is full price; cache reads are the cheapest.
By tool
Accurate spend per AI tool.
By model
Accurate spend per model.
All findings
Receipts
One receipt for every purchase of information: what you asked, each billed response, tokens, rates and amount. Click one to open, print or download it.
Claims
Mark each claim as you file it and when the provider answers. Refunded tokens are added back to your balance.
| Date | Tool | Kind | Evidence | Tokens | Amount | Status |
|---|
The meter prepares the evidence; providers decide refunds. Attach the claim report to a support request with the tool's provider.
Have I already paid for this?
Paste what you're about to ask any AI tool. If you've paid for the same request before, reuse that answer instead of buying it again.
Every billed response
One row per response after duplicates are removed. This table is also the text version of every chart.
| When | Tool | Model | Input | Output | Cache write | Cache read | Cost | Your request |
|---|
Live meter: any AI company
Send a tool's API calls through the meter and every call is recorded the moment it happens: tokens, cost, request ID, what was asked and what came back. Nothing is ever blocked, delayed or changed. When a request is identical to one already paid for, the new charge is recorded as a double spend you can claim back.
Add another AI company
Any company whose API is reached by an https address. APIs in the OpenAI, Anthropic or Gemini format are metered exactly. For any other format, calls still pass through untouched, but tokens are only recorded if the response reports usage in a format the meter recognises.
Drop usage files
CSV, JSON or JSONL usage exports from any AI tool, or Claude Code log files. You choose which column is which before anything is added.
Add one entry by hand
For tools without an export — copy the numbers from the tool's usage or billing page.
Connected sources
Remove a source to take its numbers out of every total.
Monthly budget
The overview shows how much of this month's budget is gone.
Detection
How close two requests must be to count as "already paid for".
How cost is worked out
If an import has a cost column, that number is used as-is. Otherwise the meter multiplies tokens by these prices (USD per 1 million tokens).
- Cache writes: 1.25× input price (5-minute cache), 2× (1-hour cache)
- Cache reads: the listed price, or 0.1× input when none is listed
- Built-in prices are Anthropic's first-party API rates. Add any other tool's models below.
- Subscription plans (flat monthly fee) don't bill per token; there the dollars show what the same usage would cost at API rates.
Model prices
Edit any cell. Models seen in your usage without a price are listed at the top in amber.
| Model ID | Provider | Input | Output | Cache read |
|---|