Why we bill on-demand usage at the provider's list price
No credits, no multipliers, no "fast requests". How Flintr bills usage beyond your plan, what list price means, and where our margin comes from.
Pricing4 min read
The problem with credits
Most AI tools sell usage in their own currency: credits, fast requests, premium requests, sometimes with a different multiplier per model. It looks simple on the pricing page and becomes confusing the moment you try to work out what a single agent run actually cost.
Credits also hide the price. When a provider cuts its prices, a credit system doesn't have to pass that on. When a tool changes its multipliers, your effective price changes without the headline number moving at all.
What list price means
Every model provider publishes a price per million tokens, usually split into a few kinds of token:
- Input tokens: your prompt, the files the agent reads and the conversation so far.
- Output tokens: what the model writes back, including code and explanations.
- Cached tokens: input the provider has seen recently, usually billed at a fraction of the normal input price. More on that in prompt caching, explained.
Once your included usage is used up, Flintr bills each request at exactly those published prices. When a provider lowers its price, your next request is cheaper too.
Where our margin comes from
From your subscription. Pro, Pro+ and Ultra each include a monthly amount of usage, and the subscription also pays for the editor, Tab completions, background agents and everything else we build.
That structure matters. If we made money on every token, we'd have a quiet incentive to nudge you towards more expensive models or longer agent runs. Because we don't, Auto is free to pick the cheapest model that will do the job well.
What you see in the editor
Every chat message and agent run shows the model that answered, the tokens used and the cost in dollars. The usage dashboard shows the same numbers by day and by model, and the monthly invoice has one line for on-demand usage.
Claude Sonnet · picked by Auto
18.2k input · 1.1k output · 12.4k cached
$0.07And a cap that really is hard
Transparent pricing only helps if you can stop the meter. Set a monthly cap at any amount, including $0 beyond your included usage. When you reach it, Flintr either falls back to fast models or pauses premium requests, whichever you choose. Alerts fire at the thresholds you pick, so the cap is rarely a surprise. Estimate your bill to see what a typical month looks like.
Frequently asked questions
Is there really no markup on on-demand usage?
Correct. On-demand usage is billed at the provider's published list price for each kind of token.
What happens when a provider changes its prices?
Your on-demand price changes with it, usually the same day the provider's new prices take effect.
Do Tab completions count towards usage?
No. Tab completions are unlimited on every paid plan and never draw from your included usage.
Keep reading
Pricing5 min read
Pro, Pro+ or Ultra: which plan should you pick?
A practical guide to choosing a Flintr plan, with break-even points: Pro+ becomes cheaper than Pro above $60 of monthly usage, and Ultra beats Pro+ above $210.
Read articleGuides4 min read
Prompt caching, explained: why your second agent request costs less
What prompt caching is, how model providers price cached tokens, how it shows up on your Flintr bill, and simple habits that make your agent runs cheaper.
Read articleYour editor. Your models. Your bill.
Flintr is in development and funded by the people who'll use it. Back the launch from €5, or from €25 to get into the private beta.
Coming to macOS 12+ · Windows 10+ · Linux (deb, rpm, AppImage)