Two different pricing models, two different decisions

Every cost estimate on this register shows both numbers on purpose: what a build costs with an agent subscription (Claude Pro or Max, Cursor, and similar all work this way), and what it costs paying per token through an API with no subscription. They're not the same decision dressed up two ways — they genuinely lead to different choices about what's worth building.

If you already have a subscription

The marginal cost of one more build is $0. It's already covered by a flat monthly bill you're paying regardless of whether you use it today or not. This changes the calculus completely: it's worth building small, low-stakes things you'd never justify paying $50-100 in API costs for, because there's no additional cost to weigh against the benefit. A weekend spent building a habit tracker or a QR code generator costs nothing extra beyond time, when the agent is already paid for.

If you're paying per token

Every build has a real, calculable cost, and it's worth actually calculating it before starting — this register's own cost breakdown on every app page exists for exactly this reason. Pay-per-use makes the most sense for occasional builds, or for testing whether an idea is worth pursuing before committing to a subscription. It also makes the cost of iteration visible in a way a subscription doesn't: each additional round of fixes and changes has a real, incremental price, which is a genuinely useful forcing function for writing a clearer spec upfront rather than iterating your way to one.

The actual decision rule

If you're building more than one thing a month, a subscription almost always wins on pure economics — the break-even point arrives faster than most people expect. If this is a one-off project to test an idea or replace a single subscription, pay-per-use avoids committing to an ongoing cost for something you might use once. Neither is the 'right' answer in general; they're right for different situations, and the honest move is picking based on how often you'll actually be doing this, not which one sounds more serious.