The essentials · 30-second read
n8n builds workflows and AI agents with 500+ integrations and code when needed. Self-hosted Community Edition is free; Cloud starts at €20/month billed yearly.
- Self-hosted Community Edition: free, with almost the full feature set.
- Cloud billed yearly: Starter €20/month (2,500 executions), Pro €50/month (10,000). Business €667/month, self-hosted.
- n8n counts executions (one full workflow run), not steps: long workflows cost the same as short ones.
- Our verdict: the best choice for technical teams, AI agents and data control; less friendly than Zapier for beginners.
Summarize this page with AI
n8n has become the automation tool developers reach for first, and the default choice for building AI agents that connect to real business systems. You can host it yourself for free, write code when the visual editor is not enough and pay per workflow run instead of per step. This review covers what n8n does in 2026, its verified pricing (Cloud and self-hosted), the limits of the free edition and when Zapier or Make are a better fit.
01What is n8n?
n8n (n8n.io) is a workflow automation platform created in Berlin in 2019. You build workflows on a visual canvas by connecting nodes: triggers, app integrations, logic, code and AI models. n8n now describes itself as a platform "for agents and workflows you can follow", with more than 500 native integrations and any API reachable over HTTP or MCP.

Its source code is public on GitHub (n8n says it is a top 50 GitHub project, with over 210,000 stars), under a fair-code licence: free to self-host and use internally, with restrictions on reselling it as a service. You can run it with Docker on your own server or use n8n Cloud.
02How we use n8n at BeBranded
n8n is one of the three automation tools we compare when scoping an automation project, with Zapier and Make. It is usually our recommendation when:
- The workflow includes AI. n8n's agent nodes connect a model to tools, memory and human approval steps, which is the core of most AI agent projects we scope.
- Volume is high. Paying per execution rather than per step keeps costs flat on long workflows.
- Data must stay in-house. Self-hosting keeps customer data on the client's own infrastructure.
- A developer will maintain it. Code nodes, Git (on paid plans) and environments fit a technical team.
We advise against it for a marketing team that wants to build its own simple automations without help: Zapier or Make are easier to pick up. And self-hosting is only free if someone owns updates, backups and monitoring.
03Key features
- Visual workflow editor with triggers, branching, loops, merges and error workflows.
- Code nodes: write JavaScript or Python when a built-in node is not enough.
- AI agents: connect any model, give it tools and memory, add human-in-the-loop steps, guardrails and evaluations.
- 500+ integrations, plus HTTP requests, webhooks and MCP for everything else.
- Self-hosting with Docker, on-premise or in your cloud; queue mode for scale.
- Workflow history and insights to track runs and failures.
- Enterprise controls: SSO, Git version control, environments, external secrets and log streaming on higher plans.
- MCP both ways: an MCP Server Trigger exposes your workflows as tools to AI assistants, and an MCP Client node lets an n8n agent call other MCP servers.
- Data tables and evaluations: store small datasets inside n8n, and test AI workflows with light or metric-based evaluations.
04Use cases
An AI assistant connected to company data
The most common n8n project we scope in 2026: an agent that answers questions from internal documents, a CRM or a ticketing tool. The AI Agent node connects a model to tools and memory, and n8n's human approval step lets a person validate an action (sending an email, updating a record) before it runs. Start with read-only tools and add write actions one by one.
Giving AI assistants access to your workflows through MCP
With the MCP Server Trigger, an n8n workflow becomes a tool that an AI assistant can call: "create the invoice", "look up this customer". The logic, credentials and logs stay in n8n, and the assistant only sees the tool. It is a clean way to give an assistant controlled access to internal systems. We cover the approach in our MCP service.
High-volume data processing
Syncing thousands of records between a database, a CRM and a spreadsheet is where per-execution billing pays off: a workflow that loops over 500 rows is still one execution. Watch the maximum execution duration (5 minutes on Starter) and split large jobs into batches.
Internal chatbots and forms
n8n can host or embed a chat interface and receive form submissions or webhooks, so a small internal tool (a request desk, an onboarding assistant) can run without a separate front end. When the tool gets real users and permissions, we usually move the interface to a proper internal tool.
Sensitive data on your own servers
Health, finance or HR data often cannot pass through a third-party automation cloud. Self-hosted n8n runs inside your infrastructure, next to your databases, with your own security rules. Budget the time to run it properly (see below).
05Integrations
n8n's integrations directory says "Browse 2,000+ integrations" and listed 2,757 entries when we checked, a count that mixes app nodes, trigger nodes, core nodes and partner-built nodes (the directory filters by these types). The most popular include Google Sheets, Gmail, Google Drive, Slack, Telegram, Airtable, Microsoft Excel 365, OpenAI, Google Gemini and Anthropic, plus the core AI Agent, HTTP Request and Webhook nodes.
Beyond native nodes, every plan includes custom HTTP and GraphQL requests, cURL import, webhooks, the n8n API and CLI, and code steps in JavaScript or Python. Self-hosted instances can also run custom nodes and bash scripts. On Cloud, n8n can provide AI models without your own API keys through included gateway credits on Starter and Pro.
06Getting started with n8n
- Choose Cloud or self-hosted first. Without a developer who will own the server, start on Cloud Starter; the trial needs no credit card.
- Start from a template close to your process, then replace the demo credentials with your own. Credentials are stored encrypted and can be shared on paid plans.
- Add an error workflow that alerts you (email or Slack) whenever a production workflow fails.
- Check the limits of your plan before going live: concurrent executions, maximum duration and log retention.
- For AI agents, test before you trust: run evaluations on a set of real questions and keep a human approval step on anything that writes data.
07n8n pricing in 2026
Prices checked on n8n.io/pricing on 18 November 2026 (unchanged since 14 November). The page displayed prices in euros, excluding tax, with a monthly/annual switch marked "Save 17%".

| Plan | |||||
|---|---|---|---|---|---|
| Community Edition | Billed yearly Free | Billed monthly Free | Executions per month Unlimited (your server) | Hosting Self-hosted | What you get Almost the full feature set, no projects or sharing, no SSO or Git |
| Starter | Billed yearly €20/month | Billed monthly €24/month | Executions per month 2,500 | Hosting Cloud | What you get 5 concurrent executions, 1 shared project, AI nodes and agents, forum support |
| Pro | Billed yearly €50/month | Billed monthly €60/month | Executions per month 10,000 | Hosting Cloud | What you get Up to 50 concurrent executions, 3 shared projects, admin roles, workflow history, 7 days of insights |
| Business | Billed yearly €667/month | Billed monthly €800/month | Executions per month 40,000 | Hosting Self-hosted | What you get 6 shared projects, SSO and LDAP, environments, Git version control, queue mode |
| Enterprise | Billed yearly On quote | Billed monthly On quote | Executions per month Custom | Hosting Cloud or self-hosted | What you get Unlimited projects, 200+ concurrent executions, log streaming, external secrets, SLA |
The comparison table on the same page shows the limits that matter in practice. Maximum execution duration is 5 minutes on Starter and 40 minutes on Pro and Enterprise. Saved executions go from 2,500 (Starter) to 25,000 (Pro) and 50,000 (Enterprise), execution logs are kept 7 days on Starter and 30 days on Pro, and overage fees are capped on Starter and Pro. Global variables start on Pro; environments, Git and workflow diff are Enterprise features on Cloud.
All plans include unlimited users, unlimited workflows and every integration. Starter and Pro have a free trial without a credit card; the Business trial lasts 14 days and requires one. Monthly billing costs about 20% more than the yearly price, as shown in the table. Companies under 20 employees can apply for the Startup plan (50% off Business).
Is n8n free? Yes, the self-hosted Community Edition is free and has no execution limit beyond your server's capacity. The real cost is the server and the time to maintain it: updates, backups, security and monitoring. For a small team without a developer, Cloud Starter is often cheaper in practice.
08Pros and cons
Advantages
- Billing per execution: a 20-step workflow costs the same as a 2-step one.
- Free self-hosting with almost every feature.
- The most capable tool of its kind for AI agents and complex logic.
- Code when you need it, visual editor when you do not.
- Unlimited users on every plan.
Drawbacks
- Steeper learning curve than Zapier or Make for non-developers.
- Fewer native integrations (500+) than Zapier or Make; the rest goes through HTTP requests.
- Self-hosting shifts responsibility for uptime, security and updates to you.
- Collaboration features (projects, sharing, Git) require paid plans.
- The licence restricts reselling n8n as a hosted service.
09Our verdict
Who it's for
- Technical teams and developers who want automation they can version and extend.
- Companies building AI agents connected to their CRM, documents and internal tools.
- High-volume workflows where per-step pricing would explode.
- Organisations that must keep data on their own infrastructure.
Who it's not for
- Non-technical teams that want to build alone in an afternoon: try Zapier.
- Teams with no one to maintain a self-hosted server.
- Very simple two-step automations, where n8n adds complexity for little gain.
10n8n alternatives
| Tool | |||
|---|---|---|---|
| n8n | Best for AI agents, long workflows, data control | Billing model Executions per workflow run | Self-hosting Yes, free Community Edition |
| Zapier | Best for Simple automations, widest catalogue | Billing model Tasks per action | Self-hosting No |
| Make | Best for Complex visual scenarios for non-developers | Billing model Credits per action | Self-hosting No |
| Retool Workflows | Best for Automations next to internal tools | Billing model Runs included in Retool plans | Self-hosting Enterprise only |
Our rule of thumb: n8n when a developer is involved, AI is central or data must stay home; Make when a non-technical team needs complex scenarios; Zapier for quick, simple links. Detailed comparisons: Zapier vs n8n and Make vs n8n.
11n8n vs Zapier
This is the choice most teams face. Zapier is built for non-technical users, lists more than 9,000 apps and counts a task for every action. n8n is built for technical teams, has fewer native integrations but reaches any API, and counts one execution per full workflow run, whatever the number of steps.
The cost difference grows with complexity. A ten-step workflow that runs 1,000 times a month uses 1,000 executions on n8n (within Starter's 2,500) but thousands of tasks on Zapier. For a two-step automation run a few times a day, Zapier's Free or entry plan is simpler and often enough. n8n also wins when data must stay on your servers, since Zapier cannot be self-hosted. Our detailed comparison: Zapier vs n8n.
12Self-hosting n8n: what it really takes
Running n8n yourself means a server (a small virtual machine is enough to start), Docker, a PostgreSQL database for production, HTTPS, backups and a plan for updates. You also need monitoring, because a silent failure in a workflow is worse than no automation. We set this up, or move clients to n8n Cloud when the maintenance is not worth it, as part of our automation and integration projects.
