n8n is the open-source workflow automation platform built by n8n GmbH. Visual canvas, 500+ integrations, native AI Agent nodes with multiple architectures (Tools Agent, Conversational, OpenAI Functions, Plan and Execute, ReAct, SQL), and JavaScript or Python code nodes when the visual layer runs out. Self-host for free with unlimited executions, or run managed cloud from €20/month on annual billing.
Recent developments
- May 2026: Cloud pricing restructured. Starter dropped from €24 to €20/month, Pro from €60 to €50, and Business from €800 to €667 (all annual billing). Every paid tier now includes unlimited users; tiers separate primarily on workflow-execution caps, concurrent executions, and admin features.
System Verdict
Pick n8n if the automation layer needs to be owned, not rented. Self-hosting is free with unlimited workflows and unlimited executions. The AI Agent node ships autonomous multi-step agents with tool calling across six agent architectures out of the box. JavaScript and Python code nodes close any gap the visual canvas leaves.
Skip it if onboarding speed outweighs control. Zapier ships 9,000+ integrations to n8n’s 500+. Non-technical operators land faster on Zapier or Make. Teams that refuse server ops work belong on managed platforms.
Who pays which tier: Community self-host for engineers and agencies, Starter €20/mo for small cloud use, Pro €50/mo for active production, Business €667/mo for SSO and scale, Enterprise custom for on-prem and dedicated support. All paid tiers now ship unlimited users; pricing scales on workflow-execution volume.
Key Facts
| License | Fair-code (source-available, self-host permitted) |
| Core product | Visual workflow canvas with code fallback |
| Integration count | 500+ apps |
| AI Agent node | Native · six architectures (Tools, Conversational, OpenAI Functions, Plan and Execute, ReAct, SQL) · connects to Claude Opus 4.7, OpenAI frontier models, Gemini 3.1 Pro, Mistral Large 3, Cohere, AWS Bedrock, Azure OpenAI, Ollama, and more |
| Code nodes | JavaScript · Python |
| Cloud pricing (annual) | Starter €20 · Pro €50 · Business €667 · Enterprise custom |
| Cloud users | Unlimited on every paid tier (May 2026 update) |
| Self-host | Free · unlimited workflows · unlimited executions · GitHub repo at 187k+ stars |
| Deployment | Docker · Kubernetes · VPS · Railway · any Node.js host |
| Data residency | Full control in self-host |
Every data point above was verified against vendor documentation on 2026-05-13. See Sources.
What it actually is
A workflow automation engine built on Node.js with a drag-and-drop canvas and a code escape hatch. Workflows run locally, in Docker, on a VPS, in Kubernetes, or on n8n’s managed cloud. The same workflow file moves between them.
The AI Agent node is the part worth noticing. Point it at any LLM calls with memory and tool access.
The moat is the license and the agent stack. Source-available with self-host permitted means no vendor can pull the rug. Native AI Agent nodes beat bolt-on LLM modules in Make and Zapier for agentic workloads. The trade-off is 500+ integrations to Zapier’s 9,000+ and a steeper learning curve than either.
When to pick n8n
- Data residency or compliance rules require on-prem. Self-host on a controlled server. No traffic leaves the perimeter.
- The workflow involves AI agents with tool calling. The AI Agent node is purpose-built. Make calls LLMs but does not run agent loops; Zapier Agents work but cost task credits per tool call.
- Budget is tight and volume is high. Self-host is free. A $5 to $20/month VPS handles tens of thousands of executions without cloud bills.
- The team writes code. JavaScript and Python code nodes cover anything the visual layer cannot. Credentials, versioning, and queue mode scale to production.
- Agencies run automation for clients. Self-host on the client’s infrastructure keeps margins high and data ownership clean.
When to pick something else
- Non-technical team with zero server ops: Zapier. Lower learning curve, AI Copilot builds flows from prompts, 9,000+ integrations.
- Cost-efficient complex branching without self-hosting: Make. Operations-based billing stays affordable at volume.
- LangChain-native RAG and multi-agent pipelines: Langflow chains, not SaaS glue.
- Customer-facing chat or voice agents: Voiceflow. Built for support, not workflow ops.
- Niche long-tail SaaS integration: Zapier. 9,000+ connectors covers the tail that n8n often requires HTTP nodes to reach.
Pricing
Subscription pricing via n8n.io/pricing. All cloud tiers below are quoted at annual billing.
| Plan | Monthly | Executions | Concurrent | Users | Shared projects | Who’s it for |
|---|---|---|---|---|---|---|
| Community (self-host) | $0 | Unlimited | Self-tuned | Self-managed | Self-managed | Developers and agencies land here |
| Starter (cloud) | €20 | 2,500 | 5 | Unlimited | 1 | Small teams, light cloud use |
| Pro (cloud) | €50 | 10,000 | 20 | Unlimited | 3 | Active production |
| Business (cloud) | €667 | 40,000 | Higher | Unlimited | 6 | SSO / SAML / LDAP, self-hosted option, enterprise scale |
| Enterprise | Custom | Custom | 200+ | Unlimited | Unlimited | On-prem option, dedicated support, SLA |
Prices verified 2026-05-13 via n8n pricing. All tiers now include unlimited users and every integration. Pricing is based on monthly workflow executions, not per-step cost. Qualifying startups get 50% off the Business plan. Self-hosted Community Edition has no feature caps; hosting cost is the VPS bill (typically $5 to $20/month).
Against the alternatives
| n8n | Zapier | Make | Langflow | |
|---|---|---|---|---|
| License | Source-available, self-host OK | Proprietary SaaS | Proprietary SaaS | Open source (MIT) |
| Integration count | 500+ | 9,000+ | 2,000+ | LLM-focused |
| AI agents | Native AI Agent nodes | Zapier Agents + Central | LLM modules (no agents) | LangGraph-based |
| Self-host | Free, unlimited | None | None | Free, unlimited |
| Code escape hatch | JS + Python nodes | Code step (JS/Python) | Custom functions | Custom Python nodes |
| Best viewed as | Dev-friendly open source | Incumbent generalist | Cost-efficient cloud | LLM-chain specialist |
Failure modes
- Learning curve is steeper than Zapier. The JSON data model, expression syntax, and credential system assume technical comfort. Non-technical operators stall.
- 500+ integrations lags Zapier. Niche SaaS often requires hand-built HTTP nodes and OAuth. Doable but costs engineering time.
- Self-host ops is real work. Upgrades, backups, credential encryption, queue mode, and scaling to worker processes fall on whoever runs the server.
- AI Agent node tuning matters. Complex multi-step reasoning can loop or stall without careful prompt and tool design. Not fire-and-forget.
- Cloud executions burn fast. Heavy cron schedules or high-frequency triggers chew through monthly execution limits. Self-host removes the limit entirely.
- UI clutters past 20 nodes. Large workflows become visually hard to navigate. Sub-workflows and careful grouping help.
- No enterprise-grade SaaS polish on lower cloud tiers. SSO, audit logs, and dedicated resources land only on Business and Enterprise.
Methodology
This page was produced by the aipedia.wiki editorial pipeline, an automated system that ingests vendor documentation, verifies pricing and model details against primary sources, and generates the editorial analysis you are reading. No individual human wrote this review. Scoring follows the four-dimension rubric at /about/scoring/ (Utility, Value, Moat, Longevity, unweighted average). Last verified 2026-05-13 against n8n pricing, n8n documentation, and the n8n AI Agent node docs.
FAQ
Is n8n free? Yes, self-hosted. The Community Edition runs on any Node.js host with unlimited workflows and unlimited executions. Hosting cost is the VPS bill (typically $5 to $20/month). Cloud plans start at €20/month on annual billing (n8n pricing).
What does the AI Agent node do? It runs autonomous multi-step reasoning with tool calling. The node ships six agent architectures (Tools, Conversational, OpenAI Functions, Plan and Execute, ReAct, SQL) and connects to OpenAI, Anthropic, Google Gemini, Mistral, Cohere, AWS Bedrock, Azure OpenAI, Ollama, and more. Define tools, and the agent plans and executes tasks inside a workflow (n8n Agent docs).
n8n vs Zapier? n8n wins on self-host, native AI agents, and cost at volume. Zapier wins on integration breadth (9,000+ vs 500+) and onboarding speed for non-technical users.
Is n8n really open source? Fair-code licensed: source-available, self-host permitted, commercial use permitted with limits on offering n8n itself as a managed service. The distinction matters for vendors resetting n8n as a product; it does not affect normal self-host use.
How do cloud executions count AI workflows? AI workflows count as regular executions. There is no separate AI pricing tier. A tool-calling Agent run counts as one execution regardless of how many LLM calls it makes internally.
Sources
- n8n pricing: current cloud plan rates and execution limits
- n8n documentation: self-host setup, nodes, and workflow reference
- n8n AI Agent node: native agent capabilities
- n8n GitHub: source, releases, and license
Related
- Category: AI Automation
- Comparisons: n8n vs Make vs Zapier
