Pick a plan, run studies. Top up if you need more.

Choose Starter or Team and add a card. You get up to 950 promotional credits for 7 days. You pay $0 when the trial starts; after 7 days, Stripe charges the selected plan's monthly price unless you cancel first. The plan's normal monthly allowance begins with that first paid billing period.

Studies stop until the month resets or you buy a top-up pack. Nothing rolls onto an invoice you did not agree to. Packs are $29 for 2,900 credits or $119 for 11,900 credits, they never expire, and we only draw on them once the month's allowance is gone. You can also move up a plan at any time.

Because how far a participant gets is the thing being measured, so no two studies cost exactly the same. A plan grants credits, and a study spends between 175 and 625 of them depending on how many people you send in and how far they go. The study counts on the cards are that allowance divided by a signup flow, which is the middle of the range.

Because a thirty-second first impression and a forty-step checkout audit are not the same piece of work. A flat study price would overcharge the first and lose money on the second, and we would have to cap the deep one to protect ourselves, which is exactly where the evidence gets truncated. Metering keeps the cheap study cheap.

It costs a little more, and it is usually the most valuable session in the study. Two things bound it: every participant carries a patience budget and abandons where a real user would, and every session has a cap you approve before the run. A participant cannot wander indefinitely on your credit.

It depends on where they came from. Trial credit expires 7 days after the trial starts. Your monthly allowance resets at the start of each billing period. Credits you buy in a top-up pack are yours: they never expire, and we only draw on them once the month's allowance is used up.

Every charge traces to the moments that produced it, participant by participant and stage by stage. Raw token counts are there too if you want them, they are just not the default, because nobody should need arithmetic to decide whether to run a study.

No. Every study runs on the same model configuration, because that is what lets us promise every report the same research quality. Choosing the model yourself would put those two things in tension.

No, and we would not sell it as one. Participants are AI simulations and every report says so. Use this for discovery, triage, and sharpening the questions. Take the genuinely uncertain ones to your actual customers.