Stacks and Flows

n8n

Source-available workflow automation with native AI agent nodes — self-host or use n8n Cloud.

n8n freemium Updated 2026-09-23
Visit n8n →

Use when

  • Developers and ops teams who want a workflow tool with self-hosting and code escape hatches
  • AI-heavy automations where you want to mix LLM agents and traditional integrations
  • Internal tools where data residency or self-hosting is a hard requirement

Operational upside

  • Self-host for more runtime and storage control; connected services still define the data boundary
  • Code nodes (JavaScript/Python) escape hatch when no-code hits its ceiling
  • Native AI Agent + LangChain nodes ship in core, no plugin gymnastics
  • A broad integration catalog and community nodes on a fair-code license

Tradeoffs to own

  • License is "Sustainable Use" (fair-code), not OSI-approved open source — read the fine print before commercial redistribution
  • The technical surface can be demanding for teams without an operator or developer
  • Self-hosting is "free" in license cost only; ops, upgrades, and SSO add real bills
  • Cloud pricing is based on workflow executions; workload shape determines the effective cost

What it is

n8n is a workflow automation platform that sits in the same ring as Zapier and Make.com but plays a different game: source-available, self-hostable, and built for people who are comfortable seeing — and writing — code when needed. Workflows are visual graphs of nodes; each node is a typed integration, a control structure, or a code block. Released under the fair-code “Sustainable Use License”, it can be self-hosted free for internal use, run in n8n Cloud, or embedded by ISVs under a separate commercial license.

n8n includes AI Agent nodes, LangChain integrations, vector-store nodes, and tool-calling primitives alongside its conventional automation nodes.

Who it’s for

n8n is a strong candidate for developers, indie builders, and small ops teams that want visual automation with the option to self-host. It’s especially relevant for anyone building AI agents that also need conventional integrations — query a Postgres database, post to Slack, write to S3 — because it treats LLMs as another node type rather than the center of the product.

It is a poor fit for teams that need a low-configuration managed service and do not have an owner for deployment or upgrades. Zapier and Make are relevant managed alternatives.

Strengths

  • Documented self-hosting. n8n documents Docker and npm deployment paths; production readiness still depends on the operator’s database, queue, backup, upgrade, and security choices.
  • AI-native without being AI-only. Agent and tool-calling nodes coexist with HTTP, database, and SaaS integrations. The same workflow can call Claude, hit a Postgres, send a Telegram message, and update Airtable.
  • Code escape hatch. Function nodes accept arbitrary JS or Python. When no-code runs out, you don’t have to migrate.
  • Execution-based cloud billing. n8n Cloud counts workflow executions rather than each step as a separate task; whether that costs less depends on the workflow and current plans being compared.

Public affiliate program note

n8n operates a public affiliate program. Its current terms state that approved affiliates can receive 30% on eligible n8n Cloud referrals for 12 months, subject to current program terms. This is a public program fact, not an active Stacks and Flows affiliate relationship.

Weaknesses / Watch out

  • License boundary. The Sustainable Use License is not an OSI open-source license and restricts some commercial use. Review the current text before embedding or offering a hosted service.
  • Self-hosting ownership. Identity, audit, secrets, backups, queues, and upgrades depend on the selected edition and deployment design.
  • Security patch discipline. On 16 September 2026, n8n published a coordinated security update covering credential, node, webhook, query, and approval paths. The announcement identifies fixed releases 1.123.80 for v1, 2.39.6 for the stable v2 branch, and 2.40.1 for the beta v2 branch. n8n Cloud was patched automatically; self-hosted operators below the applicable fixed release were told to upgrade. Confirm your branch and the current release notes before updating.
  • Workflow maintainability. Branching, error handling, and sub-workflow reuse should be tested by the people who will operate the result. Use the implementation guide to branching, retries, and sub-workflows when a flow outgrows linear execution.
  • Lock-in vector. Workflows are JSON-portable but the node ecosystem is n8n-specific. Migration to Activepieces or Windmill means rewriting.

Best paired with

  • Anthropic Claude or OpenAI as the LLM backend behind the AI Agent node — keep the model choice swappable rather than hard-coded.
  • Supabase or Postgres for state, vector storage, and audit logs your workflows can write to.
  • Activepieces as a comparison candidate when evaluating fully OSI-licensed alternatives.
  • Windmill when evaluating a code-first engineering workflow instead of n8n’s visual canvas; compare the two authoring models before choosing a primary surface.

Lock-in score

Medium. Workflow JSON exports cleanly and the integrations follow predictable patterns, but every n8n-specific node (the AI Agent node, the LangChain wrappers, custom community nodes) has to be rewritten when you migrate. If lock-in worries you, the practical mitigation is: keep complex business logic in code nodes (portable) rather than chains of n8n-specific nodes (not portable), and document your workflows alongside the code that calls them.

Compare n8n head-to-head

Still deciding? See how n8n stacks up against the tools it’s most often weighed against:

  • n8n vs Zapier — self-hosting and code control vs managed connector breadth
  • n8n vs Make — operator-controlled deployment vs a managed visual canvas
  • Activepieces vs n8n — license, ecosystem, deployment, and maintenance trade-offs

Frontier verification — AI nodes still run inside a workflow boundary

n8n’s AI Agent node can choose tools and generate intermediate output. The surrounding workflow still owns triggers, validation, branching, credentials, retries, and downstream writes. Calling the whole workflow an “agent” hides that useful separation.

Keep deterministic steps deterministic. If a node may retry, make external writes idempotent and record the business key used to deduplicate them.

The AI Workflow Builder can generate, refine, and debug workflows from natural-language instructions. Treat that as a build assistant, not an execution guarantee: inspect the generated nodes, credential scopes, error branches, and side effects before activation.

Current sources: n8n AI Agent node documentation, AI Workflow Builder, and n8n release notes. Verified 2026-09-14.

Verdict

Lean n8n when self-hosting and visual automation must coexist. It fits teams that want code escape hatches and AI nodes without giving up a visual workflow surface. Teams that prioritize a fully managed, non-technical experience should also evaluate Zapier or Make.

Sources

FAQ

Is n8n free?
n8n is freemium. Check the official pricing page for current tiers and limits.
What is n8n best for?
Developers and ops teams who want a workflow tool with self-hosting and code escape hatches AI-heavy automations where you want to mix LLM agents and traditional integrations Internal tools where data residency or self-hosting is a hard requirement
What are the main downsides of n8n?
License is "Sustainable Use" (fair-code), not OSI-approved open source — read the fine print before commercial redistribution The technical surface can be demanding for teams without an operator or developer Self-hosting is "free" in license cost only; ops, upgrades, and SSO add real bills
Who should use n8n?
Source-available workflow automation with native AI agent nodes — self-host or use n8n Cloud. See our review for the full pros and cons.