n8n 3.0 Breaking Changes: What Breaks, Who It Hits and How to Migrate Safely

Is n8n 3.0 already out?

Not as a final release. Since October 12, 2026, Docker Hub has an image tagged v3-rc, a release candidate meant for testing. On GitHub, the newest version on October 13 is 2.44.0, and there is no release with a 3.x number. The docs page on n8n 3.0 breaking changes nevertheless already talks about a release. Don't rely on that wording. Rely on the version number shown in your own instance.

Why the timing matters: according to the release notes, n8n ships a new minor version almost every week. The gap between release candidate and final version can therefore be short. If you prepare now, the update itself won't be a fire drill.

What is the n8n Migration Report and where do I find it?

The Migration Report checks your own instance against the changes in 3.0. You find it under Settings > Migration Report. It shows which workflows are affected and which instance settings you need to change. According to the docs for its predecessor tool, only global admins can see it.

With version 2.43 on October 6, n8n reworked the report (release notes): severity levels were replaced by four impact categories, blockers are listed first, and a progress bar shows how many workflows are already compatible.

Our take: the third row is the most dangerous one. A workflow that crashes gets noticed. A workflow that quietly does something different often goes unnoticed for weeks.

Which changes break workflows immediately?

Anything that is removed causes errors. In n8n's own words: "Workflows that use a removed node fail." These are the most common cases in instances that have grown over time:

• Function and Function Item: replace with the Code node, using the mode "Run Once for All Items" or "Run Once for Each Item"
• Item Lists: replace with the matching single-purpose node (Split Out, Aggregate, Sort, Limit, Remove Duplicates or Summarize)
• Cron and Interval: replace with the Schedule Trigger
• Read PDF, Read/Write Binary File, Convert to/from binary data: replace with Extract from File, Convert to File or Read/Write Files from Disk
• AI Agent version 1 with modes such as SQL Agent, ReAct and Conversational: move to the current version with Tools Agent
• Legacy OpenAI nodes, SerpApi, Zep, Motorhead, LangChain Code: some have a replacement, some don't; the full list is in the docs
• The $getPairedItem expression: fails at runtime, the replacement is $('Node-Name').item
• Execute Sub-workflow with the source "Local File" or "URL": import the sub-workflow and select it via "Database"

One piece of good news: n8n converts the AI Transform node into a Code node with the same JavaScript automatically during the update.

Which changes only show up later?

These changes don't produce a red error. They produce different results, or failures that only appear when a specific event happens:

1. Gmail Trigger versions 1 to 1.3 will behave like version 1.4. By default it fetches 10 emails per poll (configurable up to 50), drafts only come through with the "Include Drafts" filter, and sent or scheduled emails no longer trigger anything. If you log outgoing quotes with a Gmail Trigger, for example, those runs simply stop, without any error message.
2. Always Output Data on If and Switch: an empty item is now only created when every output is empty. Previously each empty output got one, which according to n8n made branches run that shouldn't have. Check workflows that relied on exactly that behavior.
3. Slack Trigger: the Slack credentials need a signing secret. Without it, the workflow can't be published and the trigger responds with a 401 error, so Slack events no longer arrive. You find the secret in your Slack app under Basic Information.
4. Webflow credentials (OAuth2): the "Legacy" toggle is off by default. Existing connections keep working, but the next time you reconnect, n8n requests the permissions of Webflow API v2. If you still use the old API v1, switch "Legacy" on first, otherwise Webflow rejects the connection.
5. Long-running Code nodes: the time limit for code tasks (N8N_RUNNERS_TASK_TIMEOUT) drops from 300 to 60 seconds. Longer tasks are aborted. Typical candidate: code that processes large files or many records in a loop.

What changes for self-hosted n8n?

For self-hosted instances, 3.0 is also an infrastructure update. The four changes with the biggest impact:

• Docker only: installations via npm or npx n8n are no longer supported, and n8n no longer publishes a runnable npm package. For teams and production, n8n recommends Docker Compose with Postgres instead of the bundled SQLite database.
• Community nodes: N8N_UNVERIFIED_PACKAGES_ENABLED now defaults to false. Unverified community nodes (third-party extensions) stop loading. Either set the variable to true deliberately or switch to verified nodes.
• Storage folder: ~/.n8n/binaryData is now ~/.n8n/storage. If you mounted a volume on the old path, update it. If both folders exist, n8n won't start.
• SSRF protection (a block on requests into your own network): when it is active, n8n now also blocks four additional address ranges, including 100.64.0.0/10. If your workflows call internal services in those ranges, you need to allow them explicitly.

On top of that come smaller items: the Compression node now only extracts up to 256 MiB by default instead of 2 GiB, and Chat Hub is off by default from 3.0 and will be removed entirely in n8n 4.0.

What about n8n Cloud?

On n8n Cloud, Docker, storage paths and environment variables are not your concern, but the workflow changes apply just the same. Under Updates & maintenance, the instance owner chooses the version, the release track (Beta or Stable) and the update cadence. n8n recommends Stable for business-critical workflows, and switching versions restarts the instance for one to two minutes. As of October 13, that page doesn't say how or when Cloud workspaces will be moved to 3.0. So open the Migration Report in Cloud now as well.

In what order should you migrate?

Our recommended order:

1. Back up: back up the database and export your workflows. With 3.0, n8n turns on encryption key rotation by default and explicitly advises a backup.
2. Open the Migration Report and work through the findings by category: "Upgrade blocked" and "Executions fail" first, then "Behavior changes".
3. Swap nodes while you are still on 2.x. The replacement nodes already exist there, so you can test each workflow on its own.
4. Update credentials: add the Slack signing secret, check the Legacy toggle for Webflow.
5. Test run on a copy: if you self-host, run the v3-rc image on a copy of your instance, never on production. Run your most important workflows once with real data.
6. Update to the final version, not the release candidate, and watch failed executions closely for the first few days.

What does this mean for you?

Our take: if your workflows were built in the last two years and you update regularly, the migration is usually manageable. It gets laborious for instances that have been growing since n8n 0.x or 1.x. That's where Function nodes, Cron triggers and old Gmail triggers often sit in dozens of workflows. By the way, the 1.x line still gets patches, most recently 1.123.84 on October 8. If you are still on it, 3.0 means skipping two major versions, so plan more time.

Our advice: don't update on release day, but don't wait for months either. Preparing on 2.x costs nothing and makes every workflow cleaner, and the actual update then becomes a short step. If you are still deciding whether n8n is the right tool at all, our Make vs n8n comparison has the cost breakdown. And if you connect HubSpot through n8n, keep an eye on the HubSpot private apps changes at the same time.

We handle migrations like this as part of our AI process automation: analyzing the report, rebuilding workflows, testing on a copy. If you want an estimate of the effort for your instance, get in touch.