Start free. Pay when it becomes part of how you ship.
A run costs us about two cents, so the free tier is genuinely free. Paid plans add the thing you actually buy: the recurring answer, and the alert when it changes.
Access is invite-only while we work with early users, the free tier is live now.
Snapshot
Freeone-off
See where an agent gets stuck on one site.
- One site
- Up to 3 tasks
- Evidence-backed report
- Fix-it: the fixes to paste
Starter
$29–99per month
Keep one store agent-ready.
- Everything in Snapshot
- Weekly monitoring
- Regression alerts (email + in-app)
- Renewable Agent-Ready certification
Growth
Popular$499per month
Know the day your checkout breaks for agents.
- Multi-project, daily monitoring
- Run-over-run comparison + trend
- Regression alerts
- Multi-model (GPT · Claude · Gemini)
Scale
$1,500per month
Watch a whole estate, and your rivals.
- Everything in Growth
- Recurring Competitor-watch + overtake alerts
- Actionable monitoring console
- Priority support
Business
$2.5–5Kper month
Gate every deploy on agent-readiness.
- CI gate + public API
- Team seats
- App / API / MCP surfaces
- SSO (soon)
Paid plans are opening soon, the free tier is live today. Join a plan’s waitlist above and we’ll email you the moment it opens.
Is the free tier really free?
A run costs us about two cents in model time, so the Snapshot tier stays free. It’s the honest way to let you check the tool before paying for the recurring part.
What am I actually paying for?
The evaluation is the capability; the recurring answeris the product, “still working, or here’s what changed”, with alerts when a fix breaks.
Which agents do you test with?
The models that power the assistants actually visiting sites: GPT, Claude, and Gemini, with Amazon and Perplexity coverage as an add-on. Cross-model disagreement is a finding in itself.
How is this different from Playwright?
Playwright replays a script you wrote. Frictionary sends an autonomous agent to attempt the task with no script, and shows, with evidence, where it fails.