Player not loading? Watch on YouTube
This tutorial covers a self-hosted n8n installation and its AI Assistant. The speaker starts with Docker Desktop, describes installation on Mac, Windows and Linux, and opens n8n at localhost:5678 to register an account. The supplied installer uses Docker Compose and configures a code sandbox and a private search engine. The setup message identifies it as a way to try n8n locally; the speaker points viewers to hosting documentation for further deployment details.
The walkthrough also updates an older instance. The upgrade command targets version 2.35.3, with the assistant enabled by default from 2.35. The speaker connects an OpenRouter API key and selects Claude Opus 5. This example runs n8n on the user's machine while using a model provider API. Assistant settings cover model selection, web search, MCP servers and permissions, including a control that blocks workflow deletion.
For the practical example, the assistant builds a workflow that checks a creator's page daily for new templates. It asks how to send notifications, uses a database to track previously seen templates, and requests Telegram credentials. The speaker adds a bot token and chat ID, then runs a live test. A failed notification leads to the "fix with AI" action, which sends execution context to the LLM. The speaker attributes the failure to not having started the Telegram bot and reports a successful notification afterward.
