Limits and fair usage

Free: $0, 5 enabled watches, 10,000 checks/month, minimum interval 600 seconds, 3-day event retention, 1 project, 1 member and 10 replays/day.

Builder: $19/month, 50 enabled watches, 100,000 checks/month, minimum interval 60 seconds, 7-day retention, 10 projects, 3 members and 100 replays/day.

Pro: $49/month, 200 enabled watches, 500,000 checks/month, minimum interval 60 seconds, 30-day retention, 50 projects, 10 members and 500 replays/day. Paid checkout is not available during the private beta.

Checks and budgets

Each dispatched source GET consumes one check, including 304, upstream errors, timeout, preview and manual checks. A rejected or proven undispatched request is not charged. No automatic overages. Exhaustion pauses polling until the usage period resets or a sufficient plan takes effect. Notifications are sent at 80% and 95%. Free periods are UTC calendar months; paid periods follow the subscription. A 60-second watch uses roughly 43,200 checks per 30 days; a five-minute watch uses 8,640. Interval eligibility does not mean the plan can fund every watch at that interval continuously.

Request and payload caps

HTTPS public DNS hosts, port 443, GET only, no redirects. Source: 5-second connection and 15-second overall deadline; 512 KiB compressed and decompressed caps; depth 64; selected canonical value 32 KiB. Strict JSON rejects duplicate keys, unsafe integer literals and invalid Unicode. Put large IDs in strings. Event envelope maximum 96 KiB. Metadata maximum 2 KiB. Five non-secret headers, at most 1 KiB each. Secret credential value maximum 4 KiB.

Scheduling

Scheduling runs continuously, with a five-second scan target, no catch-up bursts and positive jitter. Five concurrent source requests per workspace and two per origin; delivery has its own pool. A 429 applies a shared workspace/origin cooldown. Backoff and Retry-After can make observed intervals longer. This is polling with best-effort timing, not real-time delivery or an uptime SLA.