We want to share some big news! 
We’re launching the n8n AI Assistant, a new chat-based agent that lives right inside n8n and builds workflows for you from plain language. It can create, edit, test, and troubleshoot workflows using natural language.
The AI Assistant is available in Preview on n8n cloud from version 2.29.9
Describe what you want to automate, and the AI Assistant can plan the workflow, build it in your selected project, test it, and help you fix errors. The result is a standard n8n workflow. One you can open, edit, and publish just like any other workflow.
Running n8n self-hosted?
We’re also releasing early setup instructions for the AI Assistant on self-hosted instances. Setup is currently manual and takes some work, so it has its own post with all the details.Please keep self-hosted setup questions in that thread, and post everything else about the AI Assistant here.
What you can ask the AI Assistant to do
- Create workflows - describe the automation you want and it generates a workflow.
- Edit workflows - ask it to change logic, add nodes, or adjust configuration.
- Test and troubleshoot - ask it to run checks, inspect errors, and suggest fixes.
- And much more…
The AI Assistant has access to just about everything you do as a user, so feel free to get creative with your requests.
How to get started
-
Select AI Assistant from the left sidebar.
Make sure you’re on n8n Cloud version 2.29.9 or higher. -
Choose the Project you want it to work in.
-
Describe the task - include the trigger, the apps or services involved, the expected result, and any constraints.
-
Review the plan it proposes before it builds or changes anything.
-
Test and review the workflow before using it in production.
You stay in control
- It asks for confirmation before high-impact actions like publishing, deleting, or other important changes.
- It uses your permissions and can only access the workflows, credentials, and resources you can access in the selected project.
- Your credential and secrets are never sent to the AI. Please make sure to always use the standard n8n credential screens when entering API keys, passwords, or tokens. Never enter them in the chat.
- When web research is enabled, it asks permission before accessing any external domain.
Good to know (it’s a preview)
- The AI Assistant can make mistakes and behavior may change while it’s in development, so always review generated workflows before using them in production.
- Some capabilities may roll out gradually, and it may ask for clarification more often than usual.
- It uses credits based on the tokens processed. Longer conversations, larger workflows, and repeated iterations use more credits. If you need more credits during the Preview, you can upgrade your plan to a higher tier. We’re working on a solution to purchase additional AI credits without requiring a plan upgrade.
- The AI Assistant is not currently available on Enterprise plans.
Prompts to try
- “Create a workflow that checks Gmail every morning for invoices, saves PDF attachments to Google Drive, and adds a row to a Data Table. Ask me before publishing.”
- “Debug the latest failed execution of this workflow. Explain what failed, then suggest a fix before changing anything.”
- “Update this workflow so failed orders are sent to Slack, then retry the API request after 10 minutes.”
FAQ
Some questions you might be asking yourself …
Q. How is this different from the AI Workflow Builder?
A. The AI Assistant can do everything the workflow builder can do and more, and with better results. As such it replaces the AI Workflow Builder.
Q. How is this different from using the n8n MCP server with Claude Code (or similar)
A. In terms of capabilities, they are very similar. They both share a lot of the same underlying architecture. The AI Assistant allows you to have everything in one place and also see what the AI is building live, alongside your chat.
Q. What LLM model does it use?
A. on n8n Cloud it currently uses opus 4.8 for most of the stuff.
Q. Can I enter my own LLM API key instead of using n8n credits on n8n Cloud?
A. Not yet, but it’s something we’re thinking about.
We’d love your feedback
This is an early preview, so we’d really love your feedback to help improve the AI Assistant.
Just reply to this post with any of the following:
- What you built and how it went
- Anything that felt unclear or didn’t work as expected
- Where it got things wrong, or what you’d want it to do better
Read more in the docs: AI Assistant (Preview)
Thanks for helping us build it in the open! 