How you're charged
Token-based billing
Per token, per API call.
Duvo
Per process run.
Pricing
One fixed price per process run. No token tax.
Most AI is billed by the token, so the bill climbs the harder the model works. Duvo charges one fixed price per process run, in cents. No tokens, no seats, no surprises.
How it works
Priced by the work, not the words.
You know what a process costs before you run it. Nothing to count, nothing to license, nothing to reconcile at month-end.
You pay per process run
One invoice processed, one price change, one SKU recovered: one completed unit of work, one fixed price. Count the work, know the cost.
No tokens. No seats.
Token usage is our cost to manage, not your bill to forecast. Unlimited people can approve and monitor, and you're never charged per head.
Predictable all year
The per-process price holds whether you run a hundred or a hundred thousand. Your budget scales in a straight line, with no surprise at month-end.
How we keep it low
There's no art to throwing the biggest model at everything.
That's the expensive way, and often costs more than doing the task by hand. We engineer the cost out, so the price stays low and the work pays for itself.
Right model for the task
Small and local models handle the routine work; the frontier model is saved for the moments that actually need it. A Ferrari shouldn't idle at a toll plaza.
Sometimes the smartest AI is no AI
Where a rule, a lookup, or a line of plain code does the job, we use the line of code. Faster, exact, and never billed per token.
Cost is engineered, not passed on
Because we don't pay frontier prices for simple work, you don't either. That discipline is why a single process run costs cents.
Token billing vs. Duvo
Two ways to pay for AI. Only one you can forecast.
What drives the bill
Token-based billing
Model verbosity, context size, retries.
Duvo
The work you asked for.
One run, ten steps
Token-based billing
Costs more every step.
Duvo
Costs the same.
Seats and overages
Token-based billing
Often on top, and by surprise.
Duvo
Never. The proposal is the invoice.
Questions
Pricing, answered.
What counts as one process run?
One completed unit of work: one invoice processed, one price change, one SKU recovered. 10,000 invoices is 10,000 runs, at a price you set once and hold all year.
Do you charge per token?
No, never. Token usage is our cost to manage, not your bill to forecast. It never appears on your invoice.
Is there a per-seat fee?
No. Unlimited people on your team can approve, monitor, and review, and you're not charged per head.
What happens at high volume?
The per-process price holds. Your budget scales in a straight line with the work you run, with no surprise tiers or overage fees.
Why is the price this low?
Because we engineer the cost out. We use the right model for each task: small and local models do the routine work, plain code handles what doesn't need AI, and we only reach for a frontier model when it earns it.
How do I get an actual number?
A short call scopes your processes and expected volume. The number you agree is the number you're billed. The proposal is the invoice.