n8n MCP Server Now Builds and Updates Workflows via Natural Language

n8nn8n

n8n updated its native MCP server to allow AI agents to architect, validate, and self-heal automation workflows directly from a prompt. This shift moves n8n from a tool that agents merely trigger to a programmable environment where agents can autonomously engineer complex business logic.

n8n, an open-source workflow automation platform, launched a public preview of its native MCP (Model Context Protocol — a standard for connecting AI to tools) server's ability to build workflows. It enables n8n's Claude Code connector and ChatGPT to create and update automations from scratch using natural language.
Availability
Public Preview
Supported editions
Cloud, Enterprise, and Community
Minimum version
2.18.4
Generation method
TypeScript-based
Compatible clients
Claude Code, ChatGPT, Cursor, and more

This update allows AI agents to treat the automation platform as a programmable canvas. By generating a TypeScript representation (a structured code format for reliability) instead of raw JSON, the server ensures that agents can validate logic and perform test executions. This creates a self-healing loop where the agent fixes its own errors.

Enable the MCP server in any n8n instance running version 2.18.4 or higher, including the free Community Edition. Once connected to an MCP-compatible client, you can iterate on workflows through conversation, refining nodes and logic without manual UI configuration. The feature is currently in public preview.

View the full update on blog.n8n.io
n8n.io
n8n.io
@n8n_io
X

“Create an n8n workflow that emails me a daily weather forecast at 7am.” Minutes later: a working workflow in n8n. n8n’s MCP server can build/update workflows from Claude, ChatGPT, Cursor, or your IDE. https://t.co/Daa76USnuL

8retweets77likes
View on X

Still wondering? A few quick answers below.

The n8n MCP server is a native integration that uses the Model Context Protocol to connect external AI models directly to an n8n instance. While it previously only allowed agents to run existing workflows, the updated version enables AI assistants to create new workflows and modify existing ones using natural language instructions.

When an AI agent builds a workflow via the MCP server, it generates a TypeScript representation rather than raw JSON. This allows the agent to validate the logic, run test executions, and generate test data. If errors occur during this process, the agent can autonomously read the error logs and iterate on the code to fix the workflow.

Any AI client that supports the Model Context Protocol can connect to the n8n MCP server. This includes conversational interfaces like Claude Desktop and ChatGPT, as well as agentic coding environments like Cursor, Windsurf, and Claude Code. Users can build and update their automations directly from these external tools without switching back to the n8n user interface.

Yes, the native MCP server is a first-party feature included in every edition of the platform. This includes n8n Cloud, the Enterprise edition, and the free self-hosted Community Edition. Because it is built directly into the software, users do not need to maintain any additional third-party services or middleware to enable the connection.

To use the new workflow creation and update capabilities, users must be running n8n version 2.18.4 or higher. The feature is currently in public preview, meaning it is available for general use while the team continues to refine complex branching logic and node selection. Users must enable the server within their instance settings before connecting a client.

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

See all AI news & updates from n8n →

Share this update