Before writing a line of what became Gydmation AI, we had to decide what the AI would actually be generating. Building a proprietary workflow format was on the table — full control, no external dependency. We ruled it out for a reason that had nothing to do with engineering convenience.
A proprietary format means whatever the AI generates only runs inside our own product, forever. If we shut down, pivot, or the user just wants to leave, everything they built leaves value on the table. An open format means what gets generated is a real, ordinary n8n workflow — importable, exportable, editable by hand, runnable with or without us. That portability isn't a feature we added on top; it's a constraint we accepted before the product existed, because the alternative locks users in by design.
n8n's self-hosting story mattered as much as its openness. A business automating its own customer data, invoices, or internal communications has a legitimate reason not to want that flowing through a third party's cloud by default. Building on an engine that runs on the user's own infrastructure meant self-hosting could be the default posture of the product, not a paid add-on bolted on later.
Building on someone else's engine means inheriting its constraints — node behavior we don't control, versioning we have to track, an ecosystem that changes under us. We took that tradeoff deliberately: better to build the hard, differentiated part — turning plain English into a correct, validated workflow — on top of infrastructure already trusted at scale, than to also carry the weight of maintaining our own workflow runtime from scratch.
