Why we threw out tiered subscription plans and went pure pay-as-you-go for UnifAPI — and what that means for the way you budget agent infrastructure.

Most API marketplaces price like this: pick a tier (Basic / Pro / Ultra), pay a fixed monthly fee, get N requests included, hit a hard cap, then either upgrade or get cut off. RapidAPI is the most familiar example. It's also the most painful.

We considered doing the same thing. The financial argument for tiered subscriptions is real — it's predictable MRR, it's the dominant SaaS pattern, and it's easier to fundraise against. We threw it out anyway.

The problem with caps

Tiered plans punish the exact moment that should be celebrated: your product taking off. The day your traffic spikes is the day your provider cuts you off — or charges you 5x the per-call rate for overage. That asymmetry kills experiments and starves growth.

The cap also forces a planning fiction. You don't know how many calls you'll need this month, but you have to pick a tier as if you did. The result: you either overpay for headroom or underpay and get throttled.

What we ship instead

Free tier — a monthly credit grant that refreshes on the same date each month. No card required, all APIs unlocked, the MCP server included. It's a real tier, not a trial.

Pay-as-you-go — top up a credit balance whenever you want. Every API call costs a fixed number of credits, priced per endpoint. No monthly fee. No minimum. No hard cap. Run a single call this month or a million — the bill matches.

Enterprise — for teams who need an SLA, procurement-friendly contracts, or volume discounts. Sales-led, contract-based.

Why credits, not dollars

Credits give us a way to price each endpoint by its real cost without forcing you to think about it. A simple Twitter user lookup is cheap; an extracted-data scrape is expensive. With dollars-per-call, that requires re-publishing dozens of prices every time upstream costs move. With credits, we re-weight a single number and the user-visible price stays stable.

It's the same reason cloud platforms price in vCPU-seconds and SUs rather than raw cents per API call: it survives provider movements.

What this means for you

Budget by usage, not tier. The hardest part of agent infrastructure is knowing how much you'll consume. With pay-as-you-go, you find out by running it — and the bill matches reality. No upgrade conversations, no caps, no surprise throttling.

If your product takes off, our pricing scales with you. That's the deal.