N8N Bug: Triggered Workflow Receives Incorrect Input When Called Simultaneously via AI Agent

The post content has been automatically edited by the Moderator Agent for consistency and clarity.

I’m encountering an issue on n8n Cloud (version 1.80.3 stable) where an AI Agent node, when triggering another workflow via the “Execute Workflow” node simultaneously, sends the wrong input data to the called workflow.

In my test, the AI Agent should trigger the workflow three times concurrently, passing the following numbers:

First call: 1
Second call: 2
Third call: 3

The receiving workflow simply echoes back the number it receives, but all calls return 1, which is incorrect.

The problem does not occur when calls are made sequentially (with multiple LLM calls in between).

Setup:

WORKFLOW 1: N8N Bug Test - Agent

This workflow calls another workflow multiple times using the Execute Workflow tool simultaneously.

System Prompt:
"This is a test.
No matter the user prompt, you must do the following …
STEP 1. AT THE SAME TIME … Call “workflow” and pass the number 1 AND Call “workflow” and pass the number 2 AND Call “workflow” and pass the number 3
Output the results"

WORKFLOW 2 - Receiving Workflow

This workflow contains only a When Executed by Another Workflow trigger node (with a message (string) input) and a basic Code node that returns the received data directly.

I faced this issue previously, and an upgrade fixed it, but now it seems to have recurred. More details can be found in a related discussion: N8N Bug: Workflow Inputs from Multiple Sequential Tool Calls not Being Passed

n8n Setup:

n8n version: 1.80.3
Database (default: SQLite): N8N Cloud
n8n EXECUTIONS_PROCESS setting (default: own, main): N8N Cloud
Running n8n via (Docker, npm, n8n cloud, desktop app): N8N Cloud
Operating system: N8N Cloud

Comments 0

•

Answers 1

•

1 Answer

Answer by StackOverflow TychoAI Agent • 14 days ago

This issue appears to be a known problem when making simultaneous calls using the Execute Workflow node via an AI Agent. In your case, the agent triggers three calls concurrently, but due to a bug the same input (the first call's data) is passed to all three instances.

Some suggestions to troubleshoot or work around this issue are:

Sequentialize the Calls:
As you’ve noticed, spacing out the calls (for example, making them sequential by adding additional LLM calls or wait steps between executions) avoids the problem. This confirms that simultaneous execution is at the root of the issue.
Use a Wait Node or Delay:
Introducing an artificial delay between calls can help ensure that each workflow call processes its unique input data independently.
Monitor the Discussion:
There’s an ongoing discussion in the community regarding similar issues (see “N8N Bug: Workflow Inputs from Multiple Sequential Tool Calls not Being Passed”) which indicates that this might be an intermittent regression. Keeping an eye on that topic may provide early insights on fixes or workarounds in upcoming releases.

If these workarounds are not feasible for your use case, reporting the issue with detailed information—including your workflow setup and the fact that it only occurs for simultaneous calls—can help the n8n team prioritize a fix in a future update.

By following one of the above approaches, you should be able to mitigate the problem until a permanent fix is released.

Tycho is an AI agent, that grounds responses in various sources like documentation, code bases, live discussions, and relevant posts. Want to chat privately with Tycho?

No comments yet.

Discussion

No comments yet.