# Budgets and limits

> The monthly budgets, per-article limits and plan entitlements that keep Scout's spending inside what you agreed to, and what happens when one is reached.

Scout spends money on your behalf: model calls to judge, research and write, and paid search data to measure. Every one of those spends is checked against a limit before it happens, not reported after. When a limit is reached, Scout stops and says so. It never spends past the line and it never hides the fact that it stopped.

## The monthly budgets

Each workspace has two monthly ceilings: one for AI (the model work behind judging, research, drafting and review, and the AI visibility panel) and one for paid data (search demand, rankings, live search results). A warning appears when either reaches 80% of its ceiling. The prompt panel page shows the AI budget as a line above the questions, and turns it amber near the ceiling.

At the ceiling, scheduled work that would spend stops until the next month begins. What has already been measured stays exactly as it was. The AI visibility panel is the clearest example: it stops asking once the ceiling is reached and reports the shorter sample it took, rather than counting the questions it never asked as misses. See [Costs and cadence](/ai-visibility/costs-and-cadence).

## Per-article limits

A single article cannot run away with the budget. Each piece has a ceiling for everything Scout spends on it across research, outline, draft, review and revisions, and a cap on how many separate agent runs it may take. Each individual run has its own limits too: how many steps it may take, how many tools it may call, how long it may work. A run that reaches one of them stops and the article is marked as needing attention, with a **Try again** button on the Content page.

## Plan entitlements

The plan sets what a workspace may hold. These are the values shown in the product today.

| Entitlement                          | Limit |
| ------------------------------------ | ----- |
| Sites per workspace                  | 1     |
| Pages crawled from your site         | 300   |
| Generated articles per month         | 30    |
| Live search result checks per month  | 500   |
| Members per workspace                | 10    |
| Documents in the playbook            | 50    |
| Size of one uploaded document        | 50 MB |
| Questions on the AI visibility panel | 50    |

Where a limit is reached, the screen that would take you past it says so in place. The knowledge step of setup, for example, reads _"This workspace is holding its limit of 50 documents. Remove one to add another."_

## Manual refreshes

Some things can be asked for by hand and cost money each time, so they are rate-limited per day:

| Action                                      | Limit   |
| ------------------------------------------- | ------- |
| Opportunity refresh (a rebuild you trigger) | 3 a day |
| Re-read my site (a fresh crawl)             | 1 a day |
| Search data refresh                         | 2 a day |

When you reach one, the button explains when it will be available again. Scheduled work is not affected by these limits.

## What "stopping" looks like

Three rules hold everywhere a limit applies:

1. **Scout stops before spending, not after.** The check happens before the call.
2. **It reports the shorter sample.** A panel that asked 30 of 50 questions says 30. A scan that judged half its topics says which half.
3. **It never stops silently.** The run appears on [Activity](/automation/activity) and on Home with its reason, and anything that can be retried carries a **Retry** button.

Owners can see all of this on the workspace, and it is owners and admins who decide the rebuild cadence in [Workspace and members](/settings/workspace-and-members), the setting with the biggest effect on monthly spend.