New Self-Hosting n8n Locally + AI Assistant Setup
Summary
n8n now requires just one Docker command to self-host locally with built-in AI assistant, eliminating manual configuration. From version 2.35+, builders can bring their own LLM keys (OpenRouter, etc.) to generate entire workflows from plain English prompts, with granular permission controls preventing unwanted agent actions.
Key Takeaways
- Single-command Docker deployment eliminates manual setup of sandbox environments and private search engines that previously required hand configuration.
- AI Assistant accepts custom model keys (Claude Opus, etc.) via OpenRouter integration, letting non-technical users build workflows through conversational prompts with permission-based safety gates.
- Granular permission controls prevent AI agent from executing unwanted actions—toggle-based blocking for workflow deletion and other sensitive operations reduces security risks in multi-user instances.
- AI assistant intelligently asks clarifying questions when prompts are vague, requesting decisions on notification methods and data persistence—turning ambiguity into guided workflow design.
- Version updates use same Docker command with flags (`--update --version X.XX`), enabling one-click upgrades to access new AI features without rebuilding infrastructure.
Related topics
Transcript Excerpt
We've just simplified n8n self-hosting setup. Now, all you need to do is run this command. It sets up everything we used to configure by hand, like starting a sandbox for running code and a private search engine for the AI assistant. Speaking of which, the AI assistant comes by default from version 2.35 onwards, letting you bring your own model key to build entire workflows from plain words. Hi there, welcome to the channel. I'm Leo, and in this video, I'll be showing you how to install n8n locally, as well as how to set up the AI assistant. The first thing you'll want to do is have Docker installed. And to do so, head over to this URL, or just search for Docker Desktop in Google. And as soon as you find this page, you should be able to install it in Mac, Windows, or Linux. The installatio…