Describe the problem/error/question
I want to share something that actually happened during my class yesterday and ask for clarification.
In n8n 1.x, when using LangChain AI Agent → Call n8n Workflow, the agent could execute any existing workflow, as long as that workflow contained a trigger that AI could invoke — most commonly a Manual Trigger.
In this setup:
- The target workflow did not need to be published
- Effectively, any existing workflow could be turned into a tool
This was extremely convenient for teaching and rapid prototyping
After upgrading to n8n 2.x, I noticed a behavior change:
Regular workflows no longer appear in the “Call n8n Workflow” → from list dropdown
After investigation, it seems the following are now required:
- The called workflow must be published
- Since workflows with Manual Trigger cannot be published (because they don’t listen for events),
the trigger must be replaced with something like:- Sub-workflow trigger, or
- Webhook trigger
This creates a problem in environments where:
- The number of published workflows is limited by plan
- We previously relied on many unpublished, manual-trigger workflows as reusable AI tools
My question Is this:
an intentional policy / design change in n8n 2.x, or a current limitation with a recommended workaround?
Is there any supported way in 2.x to:
- allow AI Agents to call unpublished workflows, or
- treat Manual Trigger workflows as callable tools again?
This change has a real impact on teaching, experimentation, and AI-agent-based workflow design, so I’d really appreciate clarification from the team.
Thanks!
What is the error message (if any)?
No
Please share your workflow
Share the output returned by the last node
Information on your n8n setup
- n8n version:
- Database (default: SQLite):
- n8n EXECUTIONS_PROCESS setting (default: own, main):
- Running n8n via (Docker, npm, n8n cloud, desktop app):
- Operating system: