n8n is a workflow automation platform: connect triggers, apps and logic on a canvas, and it executes the flow. Its role on the AI side is clear — the model produces an answer, and n8n delivers it where it needs to go: into a database, a group chat, a ticket, or the next step of a process.
The 2026 Change That Matters: MCP In Both Directions
It can now play two roles at once:
- MCP server: expose existing workflows as callable tools for external AI clients.
- MCP client: let internal agents discover and call external MCP tools without hand-written API wrappers.
That matters for teams already running agents. Those automations that have "run for three years and nobody dares rewrite" can become agent capabilities as they are, without a refactor first.
It also ships an AI Agent node (model, system prompt and memory bundled together) and an AI Workflow Builder that drafts a workflow from a natural-language prompt.
Integrations and Extensibility
The company lists more than 1,800 integrations, and the node framework is open, so anything missing can be written in-house or picked up from the community. Code nodes run JavaScript or Python and can import libraries — genuinely complex logic doesn't have to be squeezed into a diagram.
Read the License
n8n is not OSI open source. It is fair-code: the source is visible, self-hostable and modifiable, but the Sustainable Use License limits three things — use and modification for internal business, non-commercial or personal purposes; redistribution only free of charge and non-commercially; and no removing licensing or copyright notices. Files with .ee. in the name fall under a separate enterprise license.
For the vast majority of teams automating their own operations, none of that bites. If you plan to wrap it as a hosted service you sell, read the license first.
Deployment and Pricing
The self-hosted community edition is free and starts with Docker. Cloud plans bill per workflow execution — one run counts once, regardless of node count or data volume — which differs materially from per-task competitors and changes the shape of the bill on heavy workflows.