n8n · Workflow and agent platform
Product and access
n8n is a workflow automation platform with an Agent Builder. Agents are distinct project artifacts and can be connected to workflows; cloud and self-hosted operation are available.
Core features
7 features from provider sources · sources reviewed on 30 September 2026. Availability and limitations are listed directly below.
Agent Builder in the project
Agents are created as their own project artifacts alongside workflows and published with versioning.
Source 3Tools, skills and knowledge base
Configurable tools, reusable Skills, and a searchable knowledge base give the agent capabilities and context.
Source 3Session and episodic memory
Sessions retain the current conversation context; episodic memory can retrieve earlier sessions.
Source 3Approval before tool calls
Selected sensitive tools may require human approval before execution.
Source 4Channels and schedules
Published agents are accessible through Slack, Telegram, or Linear and through recurring schedules.
Source 3Connect to workflows
Workflows can address agents; agents can delegate work to published subagents.
Source 3
Requirements and limits
‘Still open’ means the reviewed sources do not answer this point clearly.
Requirement
Cloud version or self-hosting
Agents are available in n8n Cloud with the current stable version; self-hosting requires additional setup and at least version 2.32.3.
Source 3Limitation
Preview phase
According to current documentation, Agents are in preview and can make mistakes; behavior may change during development.
Source 3Limitation
Self-hosted enterprise instances
According to the documentation, the agent feature is not yet available for self-hosted Enterprise.
Source 3Limitation
Queue mode
Queue mode is not yet supported for agents; n8n also notes possible errors with channel connections.
Source 3Requirement
Self-hosted knowledge base
Self-hosted knowledge bases are in preview and require a Daytona sandbox; episodic memory requires OpenAI credentials.
Source 3
Practical tasks
Editorial examples based on documented features. No product tests performed.
Prepare an incoming request
- Your input
- Form event, classification rules, and approved storage
- Possible result
- The workflow classifies the request and saves a proposed response.
- What to check
- Work with test data and error paths; control storage access.
Reply draft for approval
- Your input
- Customer question, knowledge source, and email tool
- Possible result
- The agent creates a draft; sending waits for human approval.
- What to check
- Check recipients, facts, and message before sending.
Build a request process with classification, approval, and subsequent filing.
Your possible prompt
Design a process for new requests: capture information, sort it by topic, and create a proposed response. Plan an approval step before external actions.
Getting-started idea based on documented capabilities. Not an executed product test. Adapt the task, data, and permissions.
What you should know first
n8n documentation reviewed. The AI Agent is a workflow component, not a standalone general desktop agent interface. Specific nodes, hosting options, and availability depend on the version and operating model.
Requirements & limitations
Quality depends on the workflow, data, model, and approvals. Check integrations and usage limits before use.
Setup
Configure workflow, model, and connections.
Pricing model
Cloud plan or self-hosting plus model. Check current terms and included quotas with the provider.
The classification indicates the scope of effort. It is not a price or performance commitment.
Sources from the provider
Editorial team: Michael Meierhoff · sources reviewed on . Provider information is not independent evidence of performance.
-
n8n.io /ai-agents/
-
docs.n8n.io Official website
-
docs.n8n.io /build/build-and-manage-agents.md
-
docs.n8n.io /build/integrate-ai/ai-examples/human-in-the-loop-for-tools.md
-
docs.n8n.io /choose-how-to-use-n8n/