A useful search study usually needs more than one observation. Answers, citations and organic results can change over time. ProtoScrape scheduled collections repeat a saved prompt list while keeping the timing, collection coverage and actual credit charges visible.
Start with a defined study
Choose your questions, engine and observation period before enabling automation. For example, five Google SERP questions collected once a day for seven days create at most 35 observations. At one credit per successful SERP job, the maximum successful cost is 35 credits. A failed job costs zero credits, but it is still a failed observation in your study.
The same workflow supports Google AI Mode, ChatGPT and Perplexity within the currently supported US, English, desktop market. Engine coverage varies. A scheduled observation is not a promise that an upstream answer will be complete or that it will include a citation.
Preview the calendar before saving
Open Scheduled collections, enter one prompt per line or import tagged CSV, then choose a daily or weekly local time and timezone. Preview the next five timestamps before saving. The collection runs on the server; the workspace can be closed.
During a daylight-saving change, a repeated local time runs once at its first occurrence. A nonexistent local time is skipped. The scheduler does not guarantee exact execution time. A run more than 15 minutes late is recorded as missed, and the service moves to the next future occurrence without replaying a backlog.
Separate collection allowance from actual charges
Every collection needs a lifetime run limit and a planned-credit allowance. A submitted batch allocates its full possible credit cost from that allowance, even when some jobs later fail. This bounds automatic work; it does not deduct the quote from your actual balance.
For the five-question, seven-day example, set seven run slots and a planned allowance of 35 credits. If one accepted batch has four successes and one failure, it uses one run slot and five planned credits, but charges four actual credits. A skipped occurrence uses one run slot and charges nothing.
Current account and API-key credit limits still apply. Failed, cancelled and expired jobs release their reservations. There are no automatic overage charges. Initial early-access limits are ten unarchived collections per account, ten accepted scheduled batches per account per UTC day and shared service capacity of 1,000 scheduled prompts per UTC day. Failed work remains counted toward scheduled-work limits.
Follow the outcome, including missing observations
Run history records the planned time, outcome, charged credits and a reason when collection was skipped. Each accepted run links to a normal batch, so existing result exports and signed completion webhooks work without another integration.
An unfinished previous batch blocks the next occurrence. Insufficient available credits, a spending pause, an unavailable engine or full capacity can also skip a run. Do not treat a skipped or failed observation as proof that a brand, URL or citation was absent. Report collection coverage alongside findings from completed results.
Run summaries remain for 30 days. Ordinary job results retain their existing approximate seven-day lifetime, so download them or deliver them to your receiver before expiry. A retained summary is not a retained copy of every answer.
Pause, change or end a collection
Pause stops future submissions. Any already-created batch continues and can be cancelled separately. Edit a paused collection, preview its revised timing and allowance, then resume explicitly. Editing does not reset lifetime counters. Resume selects the next future time; it does not collect missed historical answers.
Archive a finished collection to stop future runs and free a saved-collection slot. Archived configurations are scheduled for deletion after 90 days. Completed and archived collections cannot resume; create a new study when you want a new allowance.
API-key-created schedules remain attached to the creating logical key across secret rotation. Its current permissions, revocation, expiry and credit caps still apply. A signed-in owner can pause or archive a collection if that key no longer works. See the API contract for creation, preview, controls and paginated run history.