What counts as a run?
One LLM call inside the agent loop counts as one run — a discrete model request, not a token bucket. Transport retries of the same request count once, and a request that never returns a response doesn't count at all. Deterministic checks — pinned-selector actions, expects, assertions — never touch the model and never meter.
Are replays of confirmed flows free?
Yes. A confirmed flow whose pinned selectors still work replays as plain code — zero LLM calls, zero metered runs. The model re-engages only when your UI changes and a selector needs re-resolving; those resolution calls meter as runs. A CI regression test asserts this exact behavior, so the pricing claim can never silently drift from the meter.
What happens when I hit my included runs?
The monitor pauses and alerts you — it never silently rolls into overage. Overage is opt-in, per workspace, priced per run ($0.10 Starter, $0.07 Pro, $0.05 Team). Quota resets at the start of each billing period and paused monitors resume automatically.
How does the 7-day trial work?
Card upfront; the subscription auto-bills when the trial ends. Cancel before day 7 and you pay nothing. Billing is prepaid at the start of each period — there is no postpaid metering surprise.
What is a workspace?
One product. A workspace holds its sites (production, staging), monitors, flows, and quota — one subscription covers all of them up to the plan's site limit.
Can my AI agent set this up?
Yes. The CLI, HTTP API, and agent skill drive the same product humans see. An agent can register sites, run audits, and draft flows on a 7-day no-card setup tier; creating persistent monitors requires a paid plan, and MCP server access is included from Pro.