Pricing
You pay for what you send. Never a share of what it saves.
One rate, metered in bytes. No seats, no minimum, and never a share of what you save.
- $0.10per 1M tokens
- Metered exactly in bytes at $0.025 per million. Tokens are the quote, bytes are the ledger.
- 1 GBfree at signup
- $25 of credit, no card. It does not expire.
- $0everything else
- No seat licence, no monthly floor, no charge for expand().
What is actually counted
- 01
Billed on input, not output
You are billed on the payload you send us, in bytes, before compression. The output is not counted and neither is the ratio we achieve.
- 02
Bytes are the ledger, tokens are the quote
The ledger is exact in bytes because the engine has no tokenizer. The per-token price is that byte rate converted at four bytes per token — a directional figure for comparing against your model bill, not the thing we charge.
- 03
expand() never reaches us
Pure, local and free. Recovering a dropped run reads the copy already in your process.
What that comes to
| What you send | Payload | Cost |
|---|---|---|
| The free grant | 1 GB | $0 |
| A long agent session | 2 MB | $0.05 |
| A busy day of tool calls | 200 MB | $5 |
| A month of daily use | 2 GB | $50 |
Arithmetic on the list rate, not a measurement. Payload means what you send us, so compression changes the cost of your model calls, not the cost of this line.