We got tired of paying seven times.

The problem we had

By early 2025 our engineering and go-to-market teams between them held subscriptions to four AI providers, bought on four different cards, with no shared view of what any of it cost or what was being sent where. Each provider was genuinely the best at something. None was best at everything. So we kept all four, used whichever tab was already open, and stopped pretending we had a policy.

What we built instead

A thin routing layer with one login and one bill. It classified each request, sent it to the model most likely to handle it well at the lowest price, and wrote every decision to one log. Our AI spend fell by roughly half in the first month and, more usefully, became a number we could actually put in a budget.

Other teams asked to use it. That is Swurl.

What we believe

• No lab wins permanently. The best model for a given task changes every few months, and so does its price. Any architecture that assumes one provider is right forever is building in a migration.
• Cost should be visible at the moment of spending. Not at the end of the month, in an invoice, aggregated beyond usefulness. Under every message, in cents.
• Aggregation must not widen exposure. Sitting between you and eight providers is only defensible if we hold less of your data than any one of them, and we do.
• Boring beats clever in billing. Published rates, rollover, prorated changes, hard caps that actually stop. Nobody has ever been delighted by a surprising invoice.

Who we are

Swurl is owned and operated by m2Dev LLC, a software company based in Austin, Texas. We are small, we are funded by revenue rather than by a plan to become a model lab, and we intend to stay in the layer we are good at.