Six of the eight checks need no API key

Speed & errors, Re-checks and Report need no API key and keep running whatever your plan says. The two that read and reason are what you pay for.

Founding beta $0

For pointing it at one thing and seeing what comes back.

  • Speed & errors, Re-checks and Report, unlimited
  • 3 projects, 2 seats
  • Bring your own model key (Anthropic, OpenAI, Gemini, Mistral or OpenRouter), never metered, never capped
  • Includes $0 of our model usage
Start testing
Pro $29 /month

For shipping something weekly and wanting the reading agents on it.

  • Everything in Founding beta
  • $10 of model usage a month, on our key
  • 3 projects, 5 seats
  • Allowance resets monthly; credits you buy never expire
Choose Pro
Agency $149 /month

For running this across a portfolio of client products.

  • Everything in Pro
  • $30 of model usage a month, on our key
  • Unlimited projects
  • Unlimited seats
Choose Agency

What differs

Six of the eight checks are on every plan and stay there when an allowance runs out. What you buy is model usage on our key, plus room for more projects and people.

Every plan compared, feature by feature
Feature Founding beta Pro Agency
Cost
Per month $0$29$149
Model usage included, on our key $0 of model usage$10 of model usage$30 of model usage
Your own model key Never meteredNever meteredNever metered
The eight checks
Speed & errors, Re-checks, Report UnlimitedUnlimitedUnlimited
Site review and Test drive Your key onlyAllowance or your keyAllowance or your key
Standing checks kept when an allowance runs out YesYesYes
Scale
Projects 33Unlimited
Seats people in the workspace 25Unlimited
Credits you buy Not sold on this tierNever expireNever expire

How the money works

Most tools stop protecting you when the meter runs out. This one cannot, because the agents that protect you spend nothing to run.

Your key is never metered

Allowances and caps exist only for runs that would spend our API key. Point the console at your own model key and nothing here counts it, limits it, or can turn it off.

Allowances are dollars, not runs

So a bigger model simply drains faster rather than needing its own special rule, and a cheap run costs what a cheap run costs.

An exhausted allowance keeps your standing checks

Speed & errors, Re-checks and Report need no key at all. A capped account keeps every one of them, and keeps every standing check it has ever minted.

Questions before you pay

What does a run actually cost?

The reading agents cost roughly two cents a run, measured against the example product, not estimated. The deterministic ones cost exactly nothing, because they make no model call.

What is a seat, and what can each one do?

One person in a workspace. A viewer reads only and cannot start a run, because a run spends real money on somebody’s key. All four roles are written out in full.

Do I need a card to start?

No. The Founding beta tier takes no card, and the keyless checks run on it from the first minute.

Does my product’s data leave my machine?

Page text and screenshots of the product under test go to the model provider you configure, and there is no way around it. No other egress, no telemetry, no analytics.

Is my product used to train anything?

No. Your findings stay in your account, and nothing is sold, shared, or used for training.

Can I run it entirely on my own machine?

Yes, and it is the default. The console listens only on your own machine. Your findings, screenshots and key stay in two directories git already ignores.

Start with no API key

Six of the eight checks, unlimited, no key and no card. If it never finds anything worth fixing, you have lost nothing.