A collection budget starts with the number of observations you intend to make. Tracking 20 questions once is different from collecting those same questions every day for a month. ProtoScrape uses engine-weighted credits, so the engine also affects the allowance you need.
Google AI Mode currently uses five credits per successful observation. Paid subscriptions are not enabled yet; the prices shown on the website are proposed launch plans. Use the calculations below to plan workload size, and check the workspace for actual billing availability before committing to a paid workflow.
Count intended observations first
For one engine and one set of collection settings, a basic estimate is:
observations = number of questions × runs per question
credits = successful observations × credits per success
A fictional study with 20 questions, collected once a day for seven days, schedules 140 observations. If all succeed using Google AI Mode, the study uses 140 × 5 = 700 credits. If you intentionally repeat each daily observation twice, that doubles the planned credits to 1,400.
Avoid treating the question count alone as your workload. Frequency, repeats and additional supported settings can expand the total. Start with a small pilot to decide whether the extra observations answer a useful research question.
Understand available, reserved and charged credits
When the API admits a job, it reserves that job’s credits. Those credits are unavailable for another job while the first one is active. When the job succeeds, the reservation becomes a charge. Failed, cancelled or expired jobs release their reserved credits.
The workspace and GET /v1/usage show available, reserved and charged credits. Read these together when deciding whether to submit more work. A low available balance may partly reflect jobs still in progress.
Status polling and retrieval of an existing result do not create additional collection charges. Retrying a submission with the same idempotency key and identical inputs refers to the same intended job. A new key can create a new observation, so accidental duplicates can increase usage.
Translate a proposed allowance into workload
The proposed Developer plan is $29 per month for 10,000 credits, Growth is $99 for 50,000, and Scale is $299 for 200,000. At five credits per successful AI Mode result, those allowances correspond to at most 2,000, 10,000 and 40,000 successful AI Mode observations respectively, if the entire allowance is used on that engine.
Dividing a subscription price by its allowance gives an effective fully utilized credit rate. It is not a pay-as-you-go price, a checkout quote or the amount every observation necessarily costs your business. If you use only a small share of an allowance, the subscription cost per observation would be higher.
The pricing section labels these plans as proposed. Paid plans are not currently open, and automatic overage charges are not enabled. Check the current offer when billing becomes available rather than treating this guide as a purchase agreement.
Plan for concurrency and data retention too
Credits describe usage, not completion speed. Your account’s concurrency allowance limits how many jobs can be active at once, and rate limits apply to requests and submissions. Increasing the size of a study does not mean every observation can start simultaneously.
Results remain available for approximately seven days. If your report needs months of history, budget for your own storage and reporting process as well as API credits. Exporting an existing result avoids collecting the same observation again just to preserve it.
Start with the job lifecycle guide to make retries predictable, then use the workspace’s usage view to compare your plan with actual successful collection.