MCP

n8n: Build Workflows and AI Agents with a Visual Canvas and Custom Code

For building workflows and AI agents in a self-hosted environment with a visual canvas plus custom code — but you must have Docker installed and understand its fair-code license's restrictions on commercial use before your first attempt.

Official source · n8n-io/n8n

Install

curl -fsSL https://get.n8n.io | sh

Who It's For

  • Teams that want to build workflow automation and AI agents in a self-hosted environment: orchestrate processes on a visual canvas, write JavaScript or Python in nodes when needed, import npm packages, and connect to OpenAI, Anthropic, Google, or open-source models.
  • Users who want to keep both graphical building and code-level flexibility; officially described as integrating 400+ integrations.
  • When it's not a fit: n8n is distributed under a fair-code model with the Sustainable Use License and n8n Enterprise License, which is not a fully open-source license in the traditional sense. If your use case has hard licensing requirements, verify the terms yourself first.
  • The official recommendation is to run n8n via Docker; without a Docker environment, your first attempt will be blocked. Beyond that, the official prerequisites for other run methods are not stated in this evidence.

First Attempt in 3 Minutes

The recommended main path is the installation script from the official README (Docker must already be installed on your machine):

curl -fsSL https://get.n8n.io | sh
  1. Prerequisite: confirm Docker is installed and running.
  2. Run the official installation script above.
  3. If you want manual control over where data lives, the official docs also describe a manual Docker approach: first create a volume with docker volume create n8n_data, then run the docker.n8n.io/n8nio/n8n image per the official instructions and map port 5678, then open http://localhost:5678 in your browser to access the editor. Note that this is a local deployment configuration involving creating a volume and running a container locally — it is not a read-only operation. For the exact full docker run parameters, refer to the official README.
  4. Seeing the editor interface is the signal that your first attempt succeeded.

For your first session, stick to low-risk actions in the editor: browse the interface and look at the available integrations and node types. Before performing any action that connects to external services or writes data, confirm what each node does, since this evidence does not include read/write descriptions for individual nodes.

Use Cases

Case 1: Build an AI workflow. Create a workflow on the visual canvas at http://localhost:5678, insert code nodes using JavaScript or Python, and connect to OpenAI, Anthropic, Google, or open-source models. Expected outcome: the official description supports the combination of "visual building + custom code"; the exact parameter configuration of any specific model node is not stated verbatim in this evidence. The confirmed boundary is that the platform natively has AI capabilities and connectivity to these models.

Case 2: Self-hosted deployment verification. Create a data volume with docker volume create n8n_data, run the image following the manual Docker deployment steps in the official README, and visit http://localhost:5678 to verify the editor works. Expected outcome: n8n runs in local Docker and the n8n_data volume has been created; workflow execution produces local state writes, and the effects of deleting the volume or container are not covered in this evidence — verify on your own before operating.

Usage Boundaries

  • Read/write nature: the n8n editor and workflow execution trigger external service calls and local data writes (e.g., data volumes). Your first attempt should stay at the browsing and inspection level; before any real execution, be clear about what each node will do.
  • Credential requirements: the official evidence does not state whether the first launch requires account registration or an email — do not assume for unknown information.
  • Licensing restrictions: fair-code, Sustainable Use License + n8n Enterprise License; check the license terms before commercial use or redistribution.
  • Not yet confirmed: the read/write behavior of specific nodes, undo paths, data privacy boundaries, and the official prerequisites for installation methods other than Docker. None of these are covered in this evidence, which does not mean there is no risk.