💳 Costs, caps, and what uses credits

Three separate meters, no overlaps:

Model usage

Bills to your own AI subscriptions (see AI tokens). We never mark it up. Every run shows its token usage in Runs.

Credits

Pay for research tools only: web searches, page scrapes, profile lookups. 1,000 included monthly; packs of 5,000 beyond that; optional auto-recharge with a monthly cap you set. The per-tool menu lives in Billing → scrape pricing reference. Cache hits and failed calls cost zero.

Email

A flat monthly add-on per agent inbox. Sends and receives never touch credits.

Where to watch it: Home shows the week’s credit usage at a glance; Billing breaks it down by tool, agent, and mission; each run’s detail itemizes its own spend. Nothing spends silently.

The caps that protect you: email daily send caps per inbox, the auto-recharge monthly maximum, and schedule timeouts. Each is a hard stop, not a warning.

💳 Cost note: Rough anchors from real playbooks: a weekly content-research run ≈ tens of credits; watching 4 competitors twice weekly ≈ a few dozen per week; email ≈ zero credits ever. If a month looks weird, Billing → usage-by-mission usually names the culprit in one glance.