See llms.txt for all machine-readable content.

Agent, workflow, or both?

Deterministic workflows

For when nothing about your workflow is unknown in advance

  • You define every step
    Nothing is decided at runtime

  • No memory needed
    The same input gives the same result

  • Runs on a trigger
    A schedule, a webhook, an event

  • No model calls
    So no token spend at all

  • Nightly report
  • Order sync
  • Lead routing case
Why build ai agents in n8n?

You can’t control agents that you can’t understand

Agents you can audit, always

Test how your agents behave before you hit publish. Open any step to see the inputs and outputs. Open up any session, any tool call. Replay any run to find where it broke.

The predictable path to prod 

Publish in one click. Scheduling, error-handling, versioning, and execution logs come built in. Set the tools an agent can reach and actions that need a human. Self-host to keep data in-house.

Every integration is a tool

Give agents 600+ native integrations and your own workflows as tools. No need to rebuild APIs from scratch.

One company. One workspace.

Different team, different AI tool. And nobody knows what exists. In n8n, agents and workflows live in one workspace your teams can see and reuse — with access control on the ones that matter.

Trusted by teams building smarter workflows

Your token spend, 
tamed

Not every action needs an agent. So n8n doesn’t force you to use one. Mix deterministic logic with judgment only where it earns it — and stop burning AI budget on everything else.

One platform for all your 
agents and workflows

Workflows are your deterministic backbone. Agents 
add judgment and continuity on top. In n8n, they call 
each other in the same UI.

Start building

Turn plain text into 
production-ready agents

Describe it to the AI Assistant. 
Prompt it via the MCP server. Or build it by hand. n8n drafts the setup — then you test and publish.

Start building

But can’t I just build
in a coding agent?

If it’s just you, use Claude Code or Codex. Once work 
touches company systems and outlives whoever built 
it, a conversation isn’t enough.

Who depends on it?Coding agentsn8n
Just you (you watch it run) The fastest way to get it doneOverkill for a one-off task
Just you (runs while you sleep)If you wire up the scheduling, retries, and alerts yourself All come built-in to the platform
Your teamThe record lives on your machine A shared run history anyone can open and understand
Your customersWhoever runs it sets the limitsTool limits and approval steps set before anyone runs it
Ai agent template library

Where there’s a will, 
there’s already a workflow

7500+ community-built templates make getting started with agentic workflows 
simple. Even if you’re new to agents.

  • +2

Build your first AI agent

Created by: Lucas Peyrin || lucaspeyrin
Lucas Peyrin

RAG chatbot for company documents using Google Drive and Gemini

  • +9
Created by: Mihai Farcas || mihailtd

AI chatbot that can search the web

  • +4
Created by: n8n Team || n8n-team

Autonomous AI crawler

  • +13
Created by: Oskar || workfloows

Chat with a Google Sheet using AI

  • +8
Created by: David Roberts || davidn8n

AI agent that can scrape webpages

  • +7
Created by: Eduard || eduard

Personal life manager with Telegram, Google services & voice-enabled AI

  • +6
Created by: Derek Cheung || derekcheungsa
Explore templates

Reliable agents, ready for enterprise

Explore enterpriseGet started

Security + control

Fully on-prem option, SSO SAML, and LDAP, encrypted secret stores, version control, RBAC permissions.

Observability + transparency

Audit logs and log streaming to your SIEM, workflow history, real-time alerts, usage dashboards.

Developer experience

Git-based control, isolated environments, multi-user workflows, workflow diffs.

AI governance

Human-in-the-loop, guardrails, evaluations.

FAQs

Start with a prompt
Ship something that lasts

Start buildingTalk to sales