Our verdict in three lines
n8n is the builder we would hand to a small team with one technical person. One workflow run costs one execution, whether it has 3 steps or 300, so the bill stays readable as your automations grow.
Starter is $24 a month for 2,500 executions on n8n Cloud; the self-hosted Community Edition costs nothing in license fees.
The price of that freedom is a learning curve. This is low-code: you will read JSON, write the odd expression and debug your own workflows.
Good fit if
- You have someone who is comfortable with APIs, or wants to become that person
- Your workflows are long (10 steps and more) and run often
- You want the option to host everything on your own server
Look elsewhere if
- Nobody on the team wants to look at JSON: go to a no-code AI agent builder
- You need one niche app out of the box: Zapier connects 9,000+
- You plan to resell automations as a hosted service: the license restricts that
- Entry price
- From $24/mo
- Billed in
- Workflow executions
- Free plan or trial
- Free self-hosted edition
- Integrations
- 500+
- Self-hosting
- Yes
- Price checked
- Nov 15, 2026
What is n8n?
n8n (say "n-eight-n") is a workflow automation tool with a canvas: you drop nodes, connect them, and each node does one thing, such as reading a row, calling an API or asking a model. On its homepage n8n now describes itself as "the platform for agents and workflows you can follow", and the point of the product is in that last word: you can open any step and see exactly what went in and what came out.
It comes in two forms. n8n Cloud is hosted by n8n and billed monthly. The Community Edition is the same editor that you install on your own server. Both share the 500+ native integrations n8n lists on its homepage (its AI agents page says 600+), plus an HTTP node for anything with an API.
How much does n8n cost?
Prices below were read on n8n.io/pricing on November 15, 2026, with the billing toggle on "Monthly". Paying yearly lowers Starter to $20 and Pro to $50 a month.
 n8n plans9.5/10From $24/mo · checked November 15, 2026 | Monthly billing | Annual billing, per month | Included | Notes |
|---|---|---|---|---|
| Starter (Cloud) | $24 | $20 | 2,500 workflow executions a month | 1 shared project, 5 concurrent executions, agents, MCP and AI nodes, forum support |
| Pro (Cloud) | $60 | $50 | 10,000 workflow executions a month | 3 shared projects, up to 50 concurrent executions, admin roles, workflow history |
| Business (self-hosted only) | $960 | $800 | 40,000 workflow executions a month | SSO and LDAP, multiple environments, queue mode, Git version control |
| Enterprise | On request | On request | Custom executions | Contact sales; hosted by n8n or self-hosted |
Every plan includes unlimited users, unlimited workflows and every integration. What you buy is a monthly number of executions. A workflow that checks an inbox every 5 minutes runs about 8,600 to 8,900 times a month by n8n's own estimate, which already overshoots Starter. A daily report runs 30 times. Count your triggers before you count your steps.
Is n8n free?
Partly. There is no free plan on n8n Cloud, only a trial with 1,000 executions and no credit card. The self-hosted Community Edition is free to use for your own internal business purposes, and you pay for the server it runs on.
One thing to get right before you repeat "n8n is open source" in a meeting: it is not, in the strict sense. n8n uses its own Sustainable Use License, which it calls fair-code. You may use, modify and redistribute the software, but only for internal business or non-commercial purposes. Hosting n8n for paying customers needs a separate commercial agreement. The license page is here.
What can you build with n8n AI agents?
The AI Agent node sits inside a normal workflow. You give it a model, instructions, memory and a list of tools, and every other n8n integration can be one of those tools. The pricing table lists the agent features on every Cloud plan: the AI Agent node, an MCP server trigger and MCP client, hosted and embedded chat, and human approval for tool calls.
That last one matters more than it sounds. You can let an agent draft the refund and still require a person to click before money moves. n8n's own framing is a dial: fixed steps for the parts of a process you can define, an agent only where judgment is needed. It also keeps token spend down, because a deterministic step makes no model call at all.
Models are your choice. You can connect your own keys, and Cloud plans also include "AI models without API keys". The Assistant that drafts workflows from a prompt is metered separately: 1,600 Assistant credits a month on Starter, up to 9,600 on Pro.
Video: “Build your first workflow”, from the n8n channel on YouTube.
n8n Cloud or self-hosted: which one?
Take Cloud if you want to be building in ten minutes and your volume fits a plan. Take the self-hosted edition if you have data that should not leave your infrastructure, or if your execution count would make Cloud expensive. Some features only exist on a self-hosted instance (custom nodes, running bash scripts, LangSmith tracing), and the Business plan is self-hosted only.
Self-hosting is not free in practice. You need a server, backups, updates and someone who notices when a workflow fails at 2 a.m. If that person does not exist, Cloud is the cheaper option whatever the invoice says.
What are the limits of each plan?
- Starter: 2,500 executions, 5 concurrent executions, a 5-minute maximum execution time, 1 shared project, execution logs kept 7 days, forum support only.
- Pro: 10,000 executions, up to 50 concurrent executions, 40-minute executions, 3 shared projects, logs kept 30 days, admin roles and workflow history.
- Business: 40,000 executions, SSO and LDAP, environments, Git version control, queue mode. Self-hosted only, and still forum support.
- Enterprise: custom volume, audit logs, log streaming, dedicated support with an SLA.
The catch
Three things, in order of how often they bite. First, the 5-minute execution cap on Starter: a workflow that waits on a slow model or loops over a big list can hit it. Second, the jump from Pro ($60) to Business ($960) is steep, and SSO sits on the far side of it. Third, n8n rewards people who like to tinker. If the phrase "map the output of the previous node" makes you tired, you will be happier elsewhere.
What you get on day one
An editor, a library of workflow templates, data tables, the full node catalog and the agent nodes. On Cloud you also get the Assistant, which builds a first draft from a sentence. Nothing arrives pre-built for your business: you assemble it.
Hidden costs to plan for
- Model usage: with your own API keys, tokens are billed by the model provider, on top of n8n.
- Self-hosting: server, database, backups and the hours of whoever maintains it.
- Extra Assistant credits if you lean on the prompt-to-workflow feature.
- Growth: more executions means a higher tier; overage fees are listed as capped on Starter and Pro.
Questions to ask before you pay
- How many times a month will each trigger fire? That number, not the step count, picks your plan.
- Do any workflows run longer than 5 minutes?
- Do you need SSO? If yes, you are looking at Business.
- Will customers use what you build? Then read the license first.
n8n pros and cons
What we like
- Per-run billing that does not punish long workflows
- Agent node, MCP and human approval on every Cloud plan
- Self-hosted Community Edition
- Any model, including your own keys
What to watch
- Low-code: a real learning curve
- No free cloud plan, only a 1,000-execution trial
- Fair-code license, not open source
- Big price gap between Pro and Business
n8n alternatives: how it compares
The two tools closest to n8n in entry price, on the same criteria. Scores come from the same formula for all three.
| Compared |  Activepieces9.4/10From $20/mochecked Nov 15, 2026 | ||
|---|---|---|---|
| Type | Low-code, source-available | No-code, open source | No-code |
| Entry price | From $24/mo | From $20/mo | From $29.99/mo |
| Billed in | Workflow executions | Credits (one per run) | Tasks (and agent activities) |
| Free plan or trial | Free self-hosted edition | Free plan | Free plan |
| Integrations announced | 500+ | 770+ | 9,000+ |
| AI models | Connect any model with your own key; Cloud plans also include AI models without API keys | Built-in fast, smart and frontier models, or your own AI keys on Plus and up | Zapier provides the models; the AI model tier you pick changes how many tasks a step uses |
| Agents or workflows | Agents and workflows | Agents and workflows | Agents and workflows |
| Knowledge base / memory | Yes | Not stated | Yes |
| Human approval | Yes | Yes | Not stated |
| Self-hosting | Yes | Yes, limited | No |
| API / MCP | Public API, CLI, webhooks, MCP server trigger | API, SDK and MCP on every plan | Zapier MCP, SDK, webhooks on paid plans |
| Built for | Technical founders and ops people who want control and a low bill at volume | Small teams that want Zapier-style flows with a free tier and an exit to self-hosting | Non-technical teams that need an app nobody else connects |
| Official page | Check price at n8n | Check price at Activepieces | Check price at Zapier |
- n8n vs Zapier: the side-by-side, with prices checked Nov 15, 2026.
- n8n vs Make: the side-by-side, with prices checked Nov 15, 2026.
- Best AI automation tools: where n8n stands among 7 tools.
- n8n pricing: Cloud plans, executions and the free self-hosted edition
- Open these tools in the comparison tool
n8n FAQ
Is n8n really free?
The self-hosted Community Edition is free to use for internal business purposes under the Sustainable Use License; you pay for your server. n8n Cloud has no free plan, only a trial with 1,000 executions and no credit card.
Is n8n open source?
No, not under an OSI-approved license. n8n uses its Sustainable Use License, a fair-code license: the source is available and you can modify it, but commercial redistribution and hosting it for others are restricted.
What counts as an execution in n8n?
One complete run of a workflow, from trigger to last node, however many steps it contains. n8n puts it this way on its pricing page: "If it doesn't run, you don't pay."
Can n8n build AI agents?
Yes. The AI Agent node, MCP client and server trigger, hosted chat and human approval for tool calls are listed on every Cloud plan, and agents can use any n8n integration as a tool.
Sources
Everything on this page comes from n8n's own pages, read on November 15, 2026. Where a page does not state something, we say so.