n8n is a workflow automation platform. You connect triggers, apps and logic on a canvas, and it runs the flow. Its place in the AI stack is clear: the model produces an answer, and n8n delivers that answer where it needs to go — into a database, a group chat, a ticket update, the next step.
The Key Change in 2026: MCP in Both Directions
It can now play two roles at once:
- MCP server: expose existing workflows as callable tools that external AI clients use directly.
- MCP client: let internal agents discover and call external MCP tools without hand-writing a wrapper for every API.
That matters for teams already running agents: the automations that have "run for three years and nobody dares rewrite" can become agent capabilities as they are, without a refactor first.
It also offers an AI Agent node (model, system prompt and memory bundled together) and an AI Workflow Builder that drafts workflows from natural language.
Integrations and Extensibility
The official count is more than 1,800 integrations, and the node framework is open, so you can write what's missing or use community nodes. Code nodes support JavaScript and Python with third-party libraries — genuinely complex logic doesn't have to be squeezed into the visual editor.
Getting Started: Run It Locally
Following the quick start in the official repository, once Docker is installed run:
docker volume create n8n_data
docker run -it --rm --name n8n -p 5678:5678 -v n8n_data:/home/node/.n8n docker.n8n.io/n8nio/n8nThen open http://localhost:5678 in a browser to reach the editor. The n8n_data volume holds your workflows and credentials, so keep it when you upgrade the image.
A good first workflow: use a schedule trigger to pull an RSS feed or web page every morning, pass it to an AI node for a summary, and post the result to a Slack, Feishu or DingTalk group. Getting that one working teaches you the three essentials: triggers, passing data between nodes and configuring credentials.
Read the License
n8n is not open source in the OSI sense; it's what's called fair-code. The source is visible, self-hostable and modifiable, but the license (the Sustainable Use License) restricts three things: use is limited to internal business, non-commercial or personal purposes; distribution to others must be free and non-commercial; and license and copyright notices can't be removed or altered. Files with .ee. in their names fall under a separate enterprise license.
For most teams "automating our own business," these limits have no practical effect. If you plan to package it as a hosted service you sell to others, read the license first.
Deployment and Pricing
The self-hosted Community edition is free. The cloud version is billed by workflow executions — one run counts as one, regardless of node count or data volume. That differs sharply from competitors that bill per task, and with heavy flows the bill takes a completely different shape. According to the official pricing page (billed annually):
- Starter: €20 a month, 2,500 executions a month, 5 concurrent executions.
- Pro: €50 a month, 10,000 executions a month, 20 concurrent executions.
- Business: €667 a month, 40,000 executions a month, self-hosting available.
- Enterprise: custom.
Every plan includes unlimited users, unlimited workflows and every integration. Check the official site for current pricing.
Teams in China can deploy it on their own servers and connect domestic providers' APIs on the model side, keeping everything inside the network.
Tips
- Manage credentials centrally in n8n, and give each integration the least-privileged account or token possible.
- If a self-hosted instance is reachable from the internet, add HTTPS and access control rather than exposing the editor directly.
- Before letting an AI node send messages or change data on its own, add a human review step or a read-only trial run to the flow.
How It Compares
- If you don't want to self-host and the team has no technical staff: fully cloud tools such as Zapier AI are faster to start with.
- If you're building user-facing AI apps or knowledge-base Q&A rather than cross-system automation: Dify.
- If your processes revolve around spreadsheet-style data: Airtable's built-in automations may be enough.