Runs and allowance

#What a run is

A run asks every active prompt on every engine in the project once. Ten prompts on two engines make a run of twenty answers. A run is queued, worked in the background in small batches, and finished when every answer is collected or has failed.

You can leave the page: collection continues server-side. The Overview shows live progress while it's open.

#The monthly allowance

Each answer is a real, paid call to an AI provider with web search on, so each plan includes a monthly number of answers:

• It resets on the 1st of each month (UTC).
• Only answers that succeeded count. Failed answers and retries are free.
• A run that would exceed what's left isn't started; the message tells you how many remain. Pause prompts or engines, or upgrade.
• Past answers stay readable whatever your plan.

#Automatic cadence

Unless you switch automatic runs off in Project settings, Citeroot schedules runs for you. It divides your allowance by the size of one run to see how many full runs you can afford in a month, and picks the most frequent schedule that fits: daily, every two days, every three days or weekly. For example, 10 prompts × 1 engine = 10 answers per run, so Scout's 60 answers allow about six runs a month: weekly.

Automatic runs skip accounts whose email isn't confirmed (where confirmation is required) and free accounts that haven't signed in for 30 days, so abandoned accounts don't consume provider budget.

#Failures and retries

If a provider is rate-limited or briefly down, the answer is retried after a delay that grows each time (three attempts in total). If it still fails, or the provider refuses the question, it's marked failed with the reason and shown on the run.