Descriptions:
Nate Herk introduces n8n’s instance-level MCP (Model Context Protocol) feature, available from version 1.21.2 onward, which allows AI clients like Claude, ChatGPT, and Lovable to search, inspect, and execute any workflow across an entire n8n instance — not just individually pre-configured ones. This is a meaningful expansion from the earlier approach of building individual MCP server triggers per workflow.
The video walks through the complete setup process: enabling MCP access in n8n settings, connecting via OAuth to Claude’s web interface (where n8n appears as a native connector), and connecting to ChatGPT via developer mode using a custom server URL. Live demos show Claude autonomously discovering an email-sending workflow, reading its schema, and executing it without any manual endpoint configuration — then a second example updating a task manager entry by searching through available workflows to find the right one.
Herk covers the key requirements for MCP-ready workflows: they must be active or published, have the “available in MCP” toggle enabled in settings, and include clear descriptions so AI clients can understand their purpose and required inputs. He also addresses the security tradeoff — only explicitly enabled workflows are exposed, preventing unintended access to sensitive automations or API keys. The tutorial closes by briefly demonstrating how Lovable can use these same exposed workflows to build a full application. This is a practical reference for any n8n user wanting to connect their existing automation library to AI coding and chat tools.
📺 Source: Nate Herk · Published December 05, 2025
🏷️ Format: Tutorial Demo
