A Journey of a Thousand Custom Nodes

At Trade Republic, n8n isn't just a tool for wiring up webhooks — it's the backbone of how internal systems talk to each other. But once you go beyond simple HTTP requests, you start hitting the same walls over and over: discoverability breaks down, and patterns get copy-pasted across workflows — then quietly go stale as the underlying systems evolve. Worse, nobody quite remembers where the "canonical" version lived in the first place, so copy-pasting starts spreading outdated or outright bad practices instead of good ones. Add to that internal APIs that sometimes don't behave like the tidy REST services the "HTTP Request" node expects — long polling, human-in-the-loop steps, and repeated boilerplate for logging, headers, and error handling — and the cracks start to show.

In this talk, Stanislav Cherenkov, Allan Galera, and Julia Matina from Trade Republic's n8n platform team will walk through the path they took to solve that problem properly: building custom nodes at scale. They'll cover why they chose custom nodes over custom workflows, then show how they automated the boring (and error-prone) part — generating nodes directly from their internal systems' OpenAPI specs, treating each node as just another generated API client. They'll go through how that generation pipeline works, the edge cases and constraints that shaped it, and why it still holds up in practice.
Finally, they'll dig into the internal SDK built on top of this foundation: instrumentation for free (logging, tracing, env config, auth), encapsulation of repeated patterns like parameter validation and documentation generation, and a set of ergonomic conventions that make nodes easier to write and read.

Custom nodes are one of n8n's most powerful and flexible extension points, but building on them comes with real friction and general rough edges. Despite that, custom nodes remain a genuinely powerful tool — letting you shape the platform around your organization's needs without sacrificing maintenance ease or reducing the platform's capabilities.