How to Fix n8n Cannot Publish Validation Errors After the 2.43.0 Guardrails

Introduction

You finish a workflow, click Publish, and n8n refuses, even though the workflow runs fine in test mode. This is not a bug. n8n 2.43.0 added a guardrail that blocks publishing any workflow with an unconnected required node input. Here is how to find the flagged node and clear the block.

Quick answer

Connect every required input on every node, including AI sub-nodes, then save and publish again. n8n 2.43.0 blocks publishing until required inputs are connected, and no setting bypasses the check.

What broke, and why

n8n 2.43.0 shipped on October 6, 2026 as a pre-release, and its changelog carries one line that explains this error exactly: “core: Block publishing workflows with unconnected required node inputs (#37265)”. The change is confirmed in the n8n 2.43.0 release notes, and the pull request behind it is the implementation of that rule.

Before 2.43.0, you could publish a workflow with a dangling required input, for example an AI Agent node with the chat model connected but no chat trigger wired, or a sub-node that never got its input. That workflow would go live and fail in production, or fire a half-built automation against a real API. The guardrail moves the failure earlier: the editor refuses to publish until every required input is connected. It catches only structurally unconnected inputs, not wrong values, so it is a safety net, not a correctness check.

The same release added two related blocks: publishing and executing now stop when a credential attached to a node is not yours to use, and policy checks now apply to the CLI import and execute commands, so a pipeline that worked on 2.42 can fail the same way. Our n8n publish guardrails tutorial covers the wider picture.

The fix

First, look at the Publish button area: n8n shows how many nodes have issues there. Open each flagged node and the editor highlights the unconnected input or field.

Second, check the sub-nodes, not just the main nodes. The most common trap is an AI Agent node whose own inputs look fine while one of its sub-nodes, the chat model, the memory, or a tool, has an unconnected required input or a missing credential. Open every node hanging off the agent and connect what is missing.

Third, save the workflow, then click Publish again. Saving matters: n8n will not publish a workflow with unsaved changes, so an edit made after your last save is enough to keep the button blocked even after the real problem is fixed.

Pitfalls that make it look like the fix failed

A passing test run proves nothing about publishing. Manual execution runs the nodes ad hoc, while publishing runs a separate server-side validation over every node. A workflow can execute perfectly in test mode and still be refused at publish.

A second pitfall is fixing the wrong node. The flagged node is often a sub-node, not the node you were looking at. If you connect an input on the main Agent node and Publish still refuses, open the sub-nodes one by one until the issue count drops to zero.

A third pitfall is reaching for the disable key. Disabled nodes are excluded from activation validation, so disabling the flagged node makes Publish succeed. Do not ship it that way: a disabled node passes its input straight through silently, a worse failure mode than a blocked publish. Use disabling only as a one-minute diagnostic to confirm you found the culprit, then re-enable the node and fix the input.

How to verify the fix worked

Click Publish and confirm the button changes to the published state with no issue count remaining. Then trigger the workflow through its real production path, for example by sending a request to the production webhook URL, and confirm a new execution appears in the executions list. A successful publish plus one real execution is the full verification.

When this article does not apply

This fix covers the 2.43.0 block on unconnected required inputs. It does not apply if Publish spins forever with no message, which points at a hung activation path, or if the error names a credential or an organization policy instead of an input. On versions before 2.43.0 this block does not exist, so a publish failure there comes from the older validation of required fields and credentials. If every input is connected and Publish still refuses, re-check credentials and policy blocks: the check catches missing connections, not misconfiguration.

Conclusion

The 2.43.0 guardrail is n8n refusing to let a half-built workflow reach production. The fix is to finish building it: connect the missing required input, including on sub-nodes, save, and publish again. Our n8n complete guide covers the production checklist around version changes, and our fix for the 2.43.8 expression spans change covers the other 2.43-line surprise that looks like a regression but is a changed default.

Frequently asked questions