I’ve been working on this recently and realized there are quite a few small things to take care of beyond just getting the node working.
Things like the node UI, naming, descriptions, credentials, testing, packaging, publishing, and keeping everything aligned with n8n’s guidelines can add up pretty quickly.
Honestly the API call is the easy part; the edges eat the time. Credentials are usually the biggest sink - the node has to pass a credential test and fail with a readable message, and getting OAuth/header auth right (plus documenting which scopes it needs) is most of the work. Next is output shape discipline: one item in, one item out, pairedItem set so downstream expressions don’t break, and sane continueOnFail behaviour. Then the metadata nobody sees - parameter display names, descriptions, hints, and a usableAsTool description if you want the AI Agent to be able to call it. Packaging and the verification review come last, and the review will bounce you for hardcoded secrets or a missing credential test, so keep the starter repo’s structure and lint config intact from day one.
Adding to the credentials and output-shape points above, three things that save time before you ever reach review:
Scaffold with the official tool (npm create @n8n/node) and run n8n-node dev to load the node in a local n8n with live reload, instead of repackaging and reinstalling it after every change.
The same scaffold already includes a ready-made publish.yml GitHub Actions workflow. That matters because since May 1st 2026 nodes submitted for verification must be published from GitHub Actions with an npm provenance statement, and a local npm publish no longer passes. Easier to set that up on day one than to retrofit it.
Run npx @n8n/scan-community-package n8n-nodes-yourpackage early. Passing that scan is part of the verification guidelines, so it’s cheaper to catch issues on every commit than at submission time.