Free is the open-source SDK, self-hosted. Basic is the v1 hosted product: manual-first commerce with no automated processing. Managed is for merchants who outgrow the async model, and is coming soon. None of the tiers charge transaction fees.
open source · self-host
For developers and self-hosters who want full control. The SDK, no service.
@storefront/* manual-first · no automated processing
The hosted async commerce v1. You approve every order. We never touch the money.
priced accordingly
The full-service tier for merchants whose model needs real-time behavior. Everything Basic has, plus the things async excludes.
Basic is async-by-design: manual approval, manual capture, and manual fulfillment triggers. The features that need a real-time premise (automated capture, subscriptions, real-time inventory, B2B pricing tiers) live in Managed, which is coming later. We list those out up front so nobody wastes time.
These need the real-time premise that Basic doesn't carry.
Basic caps at 5,000. Beyond that you'll want Managed (coming soon), where build pipelines do incremental rebuilds and we configure the system to your scale. Or self-host with Free if you can run the build infrastructure yourself.
No, never. We don't touch the money. Your processor (Stripe, Paddle, etc.) handles payments and pays you directly. We charge a flat platform fee and that's it.
Build-time AI runs through your own provider key (OpenRouter recommended). Basic includes a small managed quota for convenience. AI output is cached on source content hash, so unchanged products don't regenerate, and a daily rebuild typically does almost no AI work.
Yes. Your catalog lives in your data store. Your customers and money live at your processor. Your code lives in your repo. We hold pseudonymous order records you can export at any time. Nothing locks you in.
We don't care. We're not the processor. Bring whatever specialty processor will work with your vertical, and we provide the catalog and orders layer. Mainstream platforms tend to refuse high-risk merchants entirely; we're agnostic.
Yes. Ollama and LM Studio work well for build-time AI when your CI runs on your own hardware. Build-time is exactly the use case local LLMs are good at: long-running, latency-tolerant, off-the-clock. Cost is zero.