Help with pricing
By size. TryCase counts the changed source files in the diff, ignoring tests, docs, lockfiles and generated code. Up to 5 files is Small ($3 with TryCase Auto, $1 with your own AI), 6–30 is Medium ($6 / $2.50), 31 or more is Large ($12 / $5). The quote appears in the pull request comment before the run starts.
A push to a pull request that has already been tested is a rerun: $1 with TryCase Auto or $0.40 with your own AI. A free pull request includes 3 reruns; later ones cost the rerun price. The diff is sized again on every push: if it has grown past what you paid for (a free slot covers up to Medium), the rerun also charges the difference to the new size, once. A commit arriving before a verdict updates the run in progress and is not charged again.
Each pull request includes 5 follow-up requests: comments or dashboard messages that steer the bot while it runs. TryCase warns you on the last free one, and each further follow-up costs the rerun price, $1 with TryCase Auto or $0.40 with your own AI, taken from the balance.
Repository setup, always. And the first 3 pull requests each calendar month (Small or Medium; Large and oversized pull requests are always paid), including their reruns that month. No card is needed for any of that. Unused free pull requests do not carry over.
You top up a prepaid balance by card, from $10 at a time, and each paid pull request is deducted from it. The balance never expires and there is no subscription. Your card is charged only when you top up yourself, unless you turn on auto top-up, which adds a fixed amount whenever the balance drops below a threshold you set.
The run waits and the pull request comment says what it costs and what the balance is. Top up on the Billing page and the run starts on its own. Earlier results and proof stay available.
If a run fails because of TryCase, for example the environment could not start, the charge goes back to your balance automatically and shows in the balance activity on the Billing page. A run that finishes with a verdict is a completed run, whether the verdict is pass or fail.
Every run gets its own environment with 4 vCPU and 8 GB of memory, with no swap. During setup, TryCase replays your recipe from a clean checkout, measures how much memory it needs with every target running, and shows the peak on the setup timeline. If a repository needs more than that, setup explains which targets to make optional or start more lightly, and the setup agent repairs the recipe automatically before giving up.
Several pull requests can be tested in parallel per account, and a bot can use up to 12 workers within a run. Your connected AI provider controls its own model limits. Need more repositories or parallel runs? Contact us about Enterprise.
Yes. The Billing page lists every top-up, charge and refund with the balance after each one, plus every pull request run with its size and price. Stripe receipts are linked from each top-up.
Contact ben@trycase.dev for help. I get back to you within 48 hours