Check What the n8n 2.43 Prerelease Changes for Publishing and Policy Checks
This page contains Coupang Partners affiliate links; the operator may earn a commission.
Key Takeaway
n8n 2.43.0 was published as a prerelease on October 6, 2026. It blocks publishing with unconnected required inputs and emits audit events on policy blocks. The stable latest is 2.42.5.

Which version is on which channel right now?
The GitHub releases list shows several versions posted between October 5 and 8.
| Version | Date | Label |
|---|---|---|
| 2.42.5 | October 8 | Stable, Latest |
| 2.43.2 | October 8 | Prerelease (beta) |
| 2.43.0 | October 6 | Prerelease |
| 2.42.3 | October 5 | Regular release |
| 1.123.84 | October 8 | 1.x line, no change items shown |
The release list alone does not tell you when the 2.43.0 changes will reach the stable channel. Please read every item below as "what is in the 2.43.0 prerelease notes".
What publishing and execution safeguards were added?
These are the items in the 2.43.0 notes related to operational safety.
- Before publishing: Publishing is blocked if required node inputs are not connected. Running and publishing are blocked if a credential the user cannot use is attached.
- During execution: Execution does not proceed if saving fails. Stop handling for queued executions and static data reloading were fixed. Engine v2 executions can be stopped from the control plane, and executions that can be suspended at node boundaries and resumed are supported.
- At shutdown: Job timeouts are capped so they do not exceed the shutdown deadline.
This makes it less likely that a half-built workflow goes out before deployment. It does not mean every error gets caught, though. It checks whether inputs are connected, which is separate from whether the values are correct.
What changed for policies and MCP?
- Policy checks apply to the CLI's import and execute commands, and when a policy blocks an action, a log streaming audit event is sent.
- MCP availability rules also apply to error workflow settings, and MCP tool telemetry attributes are redacted before they are sent.
- The notes say the one-click connector for the ChatGPT MCP client was removed. Teams that used it need to check how they connect.
- There is also an item allowing global members to grant variable scopes to API keys, so it is worth revisiting your permission design.
How are agent approvals and budgets handled?
The notes include an item saying approvals for background sub-agents are passed up to the parent conversation, an item saying an agent can mark a session as failed with a tool, and an item saying agent budget spending is stored while budget guardrails are opt-in. If guardrails are off by default, setting usage limits is up to the operator.
What should you check before upgrading?
- Keep production servers on a stable channel version, and run your publish and execution flows on the 2.43 line in a test environment first.
- Set up a log streaming destination to receive policy-block audit events.
- List whether you use anything that is changing, such as the ChatGPT MCP connector or API key variable scopes.
- Decide which version you would roll back to and where your workflow backups are.
Where do you check?
The Releases page of the official n8n GitHub repository is the reference. This article is based on the list and the 2.43.0 notes as of 2026-10-08. Some releases (such as 2.42.4) have notes that appear cut off, so open the original to confirm the exact changes.
Frequently Asked Questions
Can I put 2.43.0 on a production server now?
The release list marks 2.43.0 as a prerelease, and the latest stable version as of the same date is 2.42.5. The usual rule is to run the stable channel in production and check prereleases in a test environment first.
Are protections such as budget limits on by default?
The 2.43.0 notes say agent budget spending is stored and budget guardrails are opt-in. That means they do not apply unless you turn them on, so you need to configure them yourself.
Is anything recorded when a policy blocks an action?
The notes include an item saying a log streaming audit event is sent when a policy blocks an action. You need a log streaming destination configured to receive it.
Sources & Verification Date
- Releases (n8n-io/n8n, GitHub) · Verified
- n8n@2.43.0 release notes (n8n-io/n8n, GitHub) · Verified
All materials have been verified from official announcements and may have changed since verification.