n8n
Type: low-code · Vendor: n8n GmbH (Berlin, Germany) · Language: TypeScript · License: source-available (Sustainable Use License) · Status: active · Status in practice: mature · First released: 2019
Provide a fair-code, source-available workflow automation platform that builds event-, schedule-, or webhook-triggered workflows on a visual node canvas, with first-class AI Agent and LangChain-JS nodes for embedding LLM reasoning, RAG, and tool use inside automations.
Description. n8n is the fair-code workflow automation platform built by n8n GmbH in Berlin. Its README states: 'n8n is a workflow automation platform that gives technical teams the flexibility of code with the speed of no-code. With 400+ integrations, native AI capabilities, and a fair-code license, n8n lets you build powerful automations while maintaining full control over your data and deployments.' AI workflows are built around the AI Agent node, which the docs call 'the core of adding AI to your workflows', and the framework's AI capability is delivered via a documented LangChain-JS integration ('This feature is n8n's implementation of LangChain's JavaScript framework').
Agent loop shape. Workflows are directed graphs of typed nodes laid out in a visual editor. Each workflow starts with a trigger node (Chat, Webhook, Schedule, Form, Error, MCP Server) and pushes typed items down connections to subsequent nodes. The AI Agent root node is a cluster node requiring attached sub-nodes for chat model, optional memory, optional tools, and optional output parser. As of v1.82.0 the AI Agent always behaves as a Tools Agent (the prior 'agent type' setting was removed). Tools sub-nodes attached to the agent are looped over internally; the surrounding workflow handles routing, retries, branching, persistence.
Primary use cases
- visual workflow automation with embedded AI Agent nodes
- event-, schedule-, and webhook-triggered AI workflows
- RAG against multiple vector stores
- exposing workflows as MCP tools to external clients
Key concepts
- Workflow (docs) — Visual graph of nodes describing the automation; the unit of execution.
- Trigger node → event-driven-agent (docs) — Node that starts a workflow on an event, schedule, chat message, form submit, webhook, or upstream error.
- AI Agent node → react (docs) — Root cluster node implementing the LangChain Tools-Agent loop; requires a chat-model sub-node and at least one tool sub-node.
- Chain node (docs) — Root cluster node for non-agent LangChain chains (Basic LLM, Retrieval Q&A, Summarization, Text Classifier, Sentiment).
- Tool sub-node → tool-use (docs) — Sub-node attached to an AI Agent providing a callable tool (Calculator, HTTP Request, Custom Code, Workflow Tool, Vector Store Q&A, etc.).
- MCP Server Trigger → mcp (docs) — Trigger node that exposes n8n workflows as MCP tools to external clients over SSE or streamable HTTP.
Patterns this low-code implements —
- ★★Agentic RAG
RAG is documented end-to-end via a vector store as a tool attached to the AI Agent; vector store integrations include Simple Vector Store, PGVector, Pinecone, Qdrant, Supabase, Zep.
- ★★Approval Queue
Human-in-the-Loop middleware can require approval per tool; workflow pauses on tool call and waits for Approve/Deny on Slack/Telegram/n8n Chat.
- ★★Event-Driven Agent
Workflows start at Trigger nodes — Webhook, Chat, Form, RSS, SSE, MCP Server, Error, and many app-specific triggers.
- ★★Model Context Protocol
MCP Server Trigger exposes n8n workflows as MCP tools to external clients (SSE + streamable HTTP, no stdio). Companion MCP Client node also exists.
- ★★Scheduled Agent
Schedule Trigger runs workflows on intervals (Seconds/Minutes/Hours/Days/Weeks/Months/Custom Cron). Documented per-interval semantics with quote-verified parameters.
- ★★Structured Output
Structured Output Parser sub-node accepts a JSON schema and returns matched fields; Auto-fixing variant wraps another parser and falls back to a recovery LLM if it fails.
- ★★Supervisor
n8n ships a dedicated AI Agent Tool node that lets a root-level agent call other agents as tools — the docs explicitly call it 'multi-agent orchestration' with primary-agent supervision and delegatio…
- ★★Tool Use
AI Agent requires at least one tool sub-node; dedicated tool sub-nodes ship for Calculator, Custom Code, HTTP Request, Workflow, SerpApi, Wikipedia, Wolfram|Alpha, Vector Store Q&A.
- ★★Fallback Chain
Two indirect mechanisms: the Auto-fixing Output Parser wraps another parser and falls back to a recovery LLM if the first fails; per-workflow Error Trigger runs an error workflow when an execution fa…
Neighbourhood
Click any neighbour to follow the lineage. Scroll to zoom, drag to pan.
Alternatives & relatives
- low-code · visual builder
Flowise
Visual LangChain-JS builder; same 'visual nodes over LangChain' shape but focused on chatbot/agent flows.
- low-code · visual builder
Langflow
Visual LangChain builder (Python).
- low-code · visual builder
Dify
Visual LLM-app builder; Python-based and product-oriented vs n8n's general workflow scope.
Listed as alternative by (3)
References
Provenance
- Last analyzed:
- Last updated:
- Verification status: verified