Agents: this page is also Markdown; every page is in llms.txt.

Pricing and billing

You pay for two things: building and improving a Task, and calling it. Every build is quoted before it starts, and the quote is the most it can cost.

Plans

Plan	Price	What you get
Free	$0	5 million serving tokens a month, $10 of training credit for 30 days once an Owner adds a card, 10 drafts a month without a card
Pay as you go	Usage, billed monthly	Everything in Free; training and serving past the free amounts, charged to your card
Team	$99 a month	The same usage included, then Pay as you go rates

Drafting a Task is always free. A workspace changes plan in Settings, Billing.

v1: the Create button shows the typical cost and the cap before you start. Most first versions cost $1 to $3.
Each later change: quoted when Facet plans your Ask. A change to the Instructions alone costs cents; one that needs a new build costs about what v1 did. The quote says exactly.
The quote is a cap. A change never spends past it. If it would, it stops and asks you to raise the quote.
Stopping a change charges only what ran.

Calls are priced per million tokens, input and output, by base model:

Base model	Input, per 1M tokens	Output, per 1M tokens
Qwen3.5 0.8B	$0.04	$0.12
Qwen3.5 4B	$0.08	$0.25

Every workspace's first 5 million tokens each month are free, on any plan.

Usage is billed monthly to your card, in arrears. A month under $0.50 rolls into the next.
The Usage page shows spend by day, Task and key, and exports to CSV.
If a payment fails, an Owner pays the invoice from Settings, Billing.
One monthly spend limit per workspace stops new changes and calls when it's reached; see Limits.