How we review: this assessment is based on the product’s official documentation, current pricing and published independent evaluations; it is not a hands-on test. Read our review methodology.
n8n started as an open, self-hostable answer to cloud automation services and has grown into one of the most popular projects on GitHub, with more than 208,000 stars at the time of writing. Its pitch in 2026 is that the same canvas used to connect apps can now host AI agents with tools, memory and human approvals. This review assesses n8n as documented on its official website, pricing page and repository.
What it is
n8n is a workflow automation platform. You build workflows on a visual canvas by connecting nodes — triggers, app integrations, logic, code and AI steps — and n8n runs them on a schedule, in response to webhooks or on demand. It is available as a hosted cloud service or as software you run yourself, on-premises or in Docker. The project describes itself as “fair-code”, distributed under its Sustainable Use License and a separate n8n Enterprise License. See our n8n directory entry and our earlier Tool of the Day on n8n.
Key features
- Visual workflow builder. Drag-and-drop nodes with branching, loops and error handling, plus the option to drop into JavaScript or Python, including npm packages.
- Integrations. n8n’s homepage claims more than 500 native integrations, and any API can be reached through the HTTP Request node. (Figures elsewhere on its pages vary, from 400+ to 1,500+, depending on what is counted.)
- AI agents. Nodes for building multi-step agents with your own data, models and tools. n8n supports OpenAI, Anthropic, Google and open-source models, with human-in-the-loop approvals, structured outputs and the ability to inspect agent decisions after a run.
- MCP support. MCP nodes are included on cloud plans, and n8n says external agents and coding assistants can connect to it through MCP.
- Templates. The repository points to more than 9,000 community workflow templates.
- Enterprise features. SSO and LDAP, multiple environments, Git-based version control, queue mode for scaling, log streaming, audit logging and external secret stores on higher tiers.
Pricing
The self-hosted Community Edition is free. At the time of writing (October 2026), n8n’s pricing page lists these paid plans, priced per month on annual billing:
| Plan | Price | Executions/month | Concurrent executions | Hosting |
|---|---|---|---|---|
| Starter | €20/month | 2,500 | 5 | Cloud |
| Pro | €50/month | 10,000 | Up to 50 | Cloud |
| Business | €667/month | 40,000 | Not stated | Self-hosted |
| Enterprise | Contact sales | Custom | 200+ | Cloud or self-hosted |
All plans include unlimited active workflows. Monthly billing is available at a higher rate. An execution is one complete run of a workflow, however many steps it contains — a daily scheduled workflow costs about 30 executions a month. Companies with fewer than 20 employees can apply for a start-up plan with 50% off Business. Cloud plans include a monthly allowance of credits for n8n’s built-in AI Assistant, which do not roll over.
Strengths
Self-hosting is the defining advantage. The Community Edition runs on your own infrastructure at no licence cost, which matters for organizations that cannot send customer data, credentials or internal documents through a third-party automation service. Few comparable tools offer that option at all.
The pricing model rewards complex workflows. Many automation services bill per task or per step, so a ten-step workflow costs ten times as much as a one-step one. n8n counts the whole run as one execution, which makes long, data-heavy and AI-driven workflows considerably cheaper to operate at scale.
Developers are not boxed in. Being able to write JavaScript or Python inside a workflow, import npm packages and call any HTTP API means the visual builder rarely becomes a dead end — a common frustration with simpler no-code tools.
Its AI features are practical rather than decorative. Model choice, tool use, human approval steps and inspectable agent runs are the controls that teams evaluating agents actually need. Our guide on how to evaluate AI agents for workflow automation explains why those controls matter.
Limitations
The licence deserves attention. n8n is source-available “fair-code”, not open source in the OSI sense. The Sustainable Use License limits use to your own internal business purposes or non-commercial and personal use, and n8n’s licence FAQ restricts products whose value derives substantially from n8n functionality. Edge cases — such as building automations for clients — are covered in the FAQ but can still need a direct answer from n8n. The self-hosted Business plan also relies on a licence key that must check in with n8n’s licence server daily.
n8n asks more of its users. Concepts such as JSON data structures, expressions and webhooks are central, and non-technical users may find it harder going than consumer-oriented services.
Self-hosting shifts responsibility to you. Updates, backups, scaling, security patches and uptime become your team’s job — a real cost that the “free” label hides.
The entry-level cloud plan is modest. 2,500 executions and five concurrent runs suit small automations, but a busy chatbot workflow, where every message can count as an execution, can exhaust it quickly.
Who it’s for
- Developers and technical operations teams automating internal processes.
- Organizations with data-residency or privacy requirements that need to self-host.
- Teams building AI agents that connect models to internal tools and data.
- Agencies and consultants building automations for clients on the clients’ own infrastructure.
Alternatives
- Zapier — a gentler starting point for non-technical users, cloud-only.
- Make — a visual automation platform popular with operations and marketing teams.
- Microsoft Power Automate — a natural fit for organizations standardized on Microsoft 365.
Verdict
n8n earns 4.5 out of 5. Its documented combination of a free self-hosted edition, per-execution pricing that does not penalize complex workflows, code where you need it and well-designed AI agent controls is hard to match. We hold back half a point because the fair-code licence is more restrictive than true open source, self-hosting carries real operational costs, and the learning curve is steeper than mainstream automation tools.
