Opens in a new tab

n8n · workflow automation platform

n8n

Official documentation

n8n connects apps, data and AI models in a workflow whose steps you can inspect. A trigger starts the work, ordinary logic handles known rules, and a model can interpret input that does not fit a fixed rule. That makes it useful for processes such as ticket triage: ask the model for a team label, then validate the label before anything is routed. Tools, memory and MCP connections extend what an agent can use. You can choose managed Cloud or operate a self-hosted installation under n8n's fair-code terms. Self-hosting gives you operational responsibility; it does not stop a configured provider node from sending data elsewhere.

Product profileDocumentation checked

AutomationAgents

In this profile

When it fits

Useful for

  • Combining repeatable process logic with a bounded AI interpretation step
  • Choosing between managed hosting and an operated self-hosted workflow installation

Limits to consider

  • Assuming self-hosting prevents configured provider nodes from transmitting data
  • Using a prompt alone as the approval gate for consequential tool calls

Capabilities and platforms

  • Combines model providers, tools, memory and MCP connections in workflows
  • Mixes explicit flow logic with AI interpretation rather than delegating every step
  • Human review can gate selected tools with proposed parameters visible before a decision

Where it runs

  • Managed Cloud
  • Self-hosted software

Access and setup

Cloud account or operated self-host installation, plus separately authorized provider and app credentials. Fair-code license terms apply; provider access is distinct from n8n hosting.

Current access and pricing
  1. Choose n8n Cloud or an officially documented self-host installation and assign hosting and maintenance ownership.
  2. Create a manual-trigger workflow with synthetic input and only the provider credentials it needs.
  3. Validate the structured result and configure actual human review for any consequential tool; test errors, retries and duplicate inputs before event triggers.

Proposed exercise · not hands-on tested

Draft a support-ticket route

Input

Synthetic support tickets and an explicit list of permitted team labels.

  1. Start with a manual trigger, synthetic tickets and a fixed list of permitted teams. Ask the model to return a proposed team and a reason so each suggestion can be compared with the ticket.
  2. Validate the team against the permitted list in ordinary workflow logic. Give invalid labels, unknown cases and execution errors an explicit review path instead of silently choosing a default team.
  3. Return an internal suggestion without changing a live ticket. Repeat an input and inspect the execution history to see whether a retry would duplicate later actions; configure human review on any consequential tool added next.

Artifact to inspect

A draft routing suggestion and visible validation/error results.

What to verify

Compare each proposed route with its ticket, including an ambiguous case and an out-of-list result. Confirm that invalid labels reach review, errors remain visible and repeated inputs cannot trigger unintended duplicate actions.

Data handling and permissions

Self-hosted n8n can still transmit workflow data to whichever provider a node calls. Trace the actual authored workflow rather than applying the separate n8n assistant privacy statement to it. The privacy documentation describes assistant workflow-context handling, default telemetry and self-host deletion responsibilities, but those do not establish that an arbitrary AI workflow keeps its input local. Review provider terms, connected credentials, execution storage and retention together. In a self-hosted installation, assign updates, backups and deletion ownership as operational tasks. Use synthetic tickets for the proposed exercise and keep the model's returned text from silently acquiring authority to perform a consequential action.

Other options

Make AI Agent (New)
Compare for cloud scenario building with agent steps.
Zapier Agents / AI by Zapier
Compare for connected-app work within the Zap editor.

Sources and scope

Documentation-based product profile; no hands-on testing. The proposed workflow is an authored exercise, not a verified outcome or a practical guide. Capabilities and policy statements describe vendor documentation checked October 2, 2026; task-fit comparisons are editorial judgments. Plan, region, administrator and interface limits may apply.

Source review: . Product access and terms can change.

Suggest a correction