In this video, I’m breaking down why I still build n8n AI agents even though Claude Code exists—and honestly, it’s because the combo is way more powerful than choosing one or the other. n8n just released MCPs at the instance level, which means Claude can take actions directly inside your n8n without you constantly jumping back to the dashboard. I show you how the “connector” approach works, where Claude can find and run workflows, but then I take it a step further: using MCP servers + Claude Code to actually build entire workflows for you. I walk you through setting up a project in VS Code, creating an MCP.json file, and wiring it to your n8n API URL + API key. Then I install two community repos from CZ Lowski (shout out): n8n-mcp (a more Claude-friendly MCP than the official one) and n8n-skills (Claude Skills that teach Claude how to structure real automations). After that, I copy an n8n template idea (AI content factory from RSS → AI filtering → parallel content generation → Slack approval → Google Sheets) and let Claude build it inside my n8n instance. It hits a few validation errors, but Claude fixes them automatically, and in the end I get a working workflow where 70–80% of the job is done—then I just add credentials and activate.