Tested by SkillHammer Read-only scanner (Python) + Agent Skill for Claude Code, Cursor or Codex

n8n 3.0 Upgrade Check

Find what n8n 3.0 will break in your workflows and deployment files before an upgrade you can't roll back.

n8n 3.0 removes 37 node types (Function, Cron, Item Lists, legacy OpenAI and more), old AI Agent modes, $getPairedItem and $evaluateExpression() in Code nodes, and npm installs: self-hosting is Docker-only. And 3.x can't be rolled back to 2.x.

This check runs on files, not a running instance: workflow exports, .env, docker-compose.yml, package.json, systemd and PM2 files. So it also works for old 1.x and 2.x installs, backups, client audits and CI.

25 checks (11 blockers, 9 warnings, 5 notes), each with the exact fix. Markdown report for a ticket, JSON and exit codes for CI. One Python file, no dependencies, runs offline.

What's in the download

• scripts/n8n3_scan.py: the scanner
• SKILL.md: an Agent Skill that runs the scan and walks you through the fixes
• reference/BREAKING-CHANGES-3.0.md: every change, with sources
• tests/ with fixtures, TEST-REPORT.md, README, LICENSE.txt, REFUND.txt

Who it's for

People running self-hosted n8n who are planning the move to 3.0, and consultants auditing client instances or backups.

What's tested

• 24/24 tests: 17 scanner, 3 cross-checks against the real n8n npm packages, 2 on a real n8n 2.45.7 and 2 on a real n8n 3.0.1
• The removed-node list equals the npm diff exactly: 572 node types in 2.45.6, 535 in 3.0.1, 37 removed
• On real n8n 3.0.1 (the official image run in a chroot, not Docker), 9 probe workflows ran or failed exactly as the scanner predicted

What isn't tested

• A real Docker or Kubernetes deployment, and n8n's own Migration Report UI
• Real customer exports (fixtures are hand-written with real node types and versions)
• Settings that only exist in a running instance; these are listed in SKILL.md Limits
• Windows paths and Python versions below 3.13

Read the full TEST-REPORT

Requirements

Python 3.9+ (tested on 3.13), standard library only. Your n8n workflow exports and deployment files.

How to use it

Export your workflows (n8n export:workflow --all --separate), then run python3 scripts/n8n3_scan.py <exports> --deploy <folder with .env/compose> --markdown.