n8n
n8n is a workflow automation platform for technical teams that pairs a visual canvas with real JavaScript or Python code. It connects 500+ apps, builds traceable AI agents, and runs on your own infrastructure or n8n's cloud.
What is n8n?
n8n is a workflow automation platform published by n8n GmbH in Berlin and aimed squarely at technical teams. Its interface is a canvas on which every step of a process becomes a node, but unlike most no-code tools it never locks the builder out of the code: a Code node runs JavaScript or Python anywhere in the flow, dynamic expressions parameterise each step, and self-hosted instances can pull in npm packages. The company states the trade-off it refuses to make — other tools limit you to either a visual building experience, or code.
The catalogue advertises more than 500 integrations, with over 400 pre-configured nodes covering messaging, CRM, databases, spreadsheets, storage and cloud services; an HTTP Request node and a GraphQL node handle everything else, and cURL commands can be imported directly. On the AI side, a family of LangChain-based nodes provides agents, chains, memory, embeddings, vector stores, retrievers, output parsers, text splitters, rerankers and Model Context Protocol support, so multi-agent set-ups and RAG pipelines are assembled on the same canvas as the rest. Models can be cloud-hosted or run locally, and every step of an agent's reasoning stays visible on the canvas. Structured inputs and outputs, human-in-the-loop approvals and native evaluations keep AI actions bounded.
Debugging absorbs much of the design effort: a single step can be re-run without replaying the whole workflow, data can be mocked or replayed instead of re-triggering external systems, and an interleaved log view shows the output of each line.
Deployment is the other structural choice. n8n runs under Docker on your own servers — the full source code sits on GitHub under a fair-code licence, with more than 200,000 stars — or on n8n's hosted cloud. Enterprise governance covers SSO with SAML and LDAP, encrypted secret stores, RBAC, audit logging, log streaming to a SIEM, Git-based version control, isolated environments and workflow diffs. More than 10,000 templates, a public REST API, a CLI and a community forum of over 200,000 members surround the product. Huel, Vodafone and Musixmatch are named as customers.
What it does
- Connect over 500 applications, plus any other API through the HTTP Request node
- Build multi-step AI agents and RAG systems with LangChain-based nodes
- Plug in the model of your choice, cloud-hosted or local, using your own API keys
- Write JavaScript or Python at any point in a workflow
- Trigger runs from webhooks, schedules, chat, application events or another workflow
- Insert human approvals and guardrails around AI decisions
- Deploy on your own infrastructure with Docker, or on n8n's hosted cloud
When to use n8n / When not to
A quick filter to help you decide if n8n is the right fit.
When to use n8n
- IT operations teams automating repetitive internal processes such as employee on-boarding
- Security operations analysts enriching incident tickets and building SOAR-style playbooks
- Developers and platform engineers who want a visual canvas without giving up JavaScript or Python
- Data and AI engineers assembling multi-agent systems, RAG pipelines and model-agnostic workflows
- Organisations with data-residency or sovereignty constraints that need a fully on-premise deployment
When not to use n8n
- Non-technical users looking for a point-and-click assistant: the canvas assumes APIs, credentials, webhooks and JSON
- Teams that need guaranteed response times, since dedicated support with an SLA is reserved for the Enterprise plan
- Agencies or resellers wanting to resell the service, as the self-serve terms restrict use to internal business purposes
- Mobile-first users: there is no iOS or Android application, only a web interface and a self-hosted instance
- High-risk contexts where an error could cause death, serious injury or significant legal liability, which the AI terms exclude
How to use n8n
A typical end-to-end flow, from setup to results.
- Choose your deployment: create a cloud account, or self-host from GitHub with Docker following the hosting documentation
- Start a free trial on Starter or Pro without entering a credit card, or apply a licence key on a self-hosted instance for Business and Enterprise features
- Open the canvas and place a trigger: a webhook, a schedule, a chat message, an application event or another workflow
- Connect the services you need, preferring OAuth where the provider supports it and otherwise supplying a scoped API key
- Chain the nodes that move and transform the data, using switches, filters, loops and merges
- Drop in a Code node in JavaScript or Python, or a dynamic expression, wherever the pre-built nodes stop short
- Add AI nodes to build an agent or a RAG pipeline, pointing them at a cloud or local model with your own credentials
- Test as you go: run a single step, replay or mock data, and read the interleaved log view
- Add human-in-the-loop approvals, guardrails and error workflows before anything reaches production
- Promote to production with Git version control, separate environments, execution alerts and the insights dashboard
Pros & Cons
Pros
- Full self-hosting under Docker with the source code published on GitHub, so data need never leave your infrastructure
- Visual building and real code in the same tool, with no context switch between them
- Billing counts complete executions rather than steps or seats, and every plan includes unlimited users, workflows and integrations
- Written commitment in the AI terms not to train machine-learning models on Customer Content or Output
- Cloud data hosted in the European Union, on servers in Frankfurt, with a named list of sub-processors published
- Security programme aligned to SOC 2 with annual independent audits, and a SOC 3 report available for download
- Large ecosystem: 10,000+ workflow templates, a forum of over 200,000 members, free courses and 200k GitHub stars
Cons
- Steep jump in the price ladder between Pro at 50 EUR per month and Business at 667 EUR per month
- No permanent free plan is advertised in the cloud pricing grid, and the self-hosted Community Edition is presented without any mention of price
- Dedicated support with an SLA is Enterprise-only; the Business plan is explicitly self-serve and leans on the community forum
- Real technical prerequisites: APIs, credentials, webhooks and JSON structures are unavoidable
- Self-hosting shifts encryption at rest and TLS termination onto the user
- Annual commitments are non-refundable, with no pro-rata refund, and Business overages are billed at 4,000 EUR per extra 300,000 executions
- The AI Assistant is still a preview with monthly credits that do not roll over and cannot be bought separately
Pricing & Plans
There is no permanent free plan advertised on the cloud pricing grid: all four published tiers are paid, and the self-hosted Community Edition is offered on GitHub without any stated price. A free trial is available on the Starter and Pro plans with no credit card required, and on the Business plan for fourteen days with a card. The lowest paid entry point is the Starter plan at 20 EUR per month when billed annually, which includes 2,500 monthly workflow executions, unlimited users and workflows, and every integration. Annual billing is advertised as a 17% saving over monthly billing.
- 2
- 500 workflow executions
- 1 shared project
- 5 concurrent executions
- unlimited users
- 2
- 300 AI credits per month
- forum support
- 10
- 000 workflow executions
- 3 shared projects
- 20 concurrent executions
- 7 days of insights
- up to 13
- 700 AI credits per month
- admin roles
- 40
- 000 workflow executions
- self-hosted
- 6 shared projects
- SSO SAML and LDAP
- 30 days of insights
- separate environments
- scaling options
- custom execution volume
- hosted by n8n or self-hosted
- unlimited shared projects
- 200+ concurrent executions
- 365 days of insights
- external secret store
- log streaming
- extended data retention
- Start-up Plan — 50% off the Business plan for companies under 20 employees
- subject to eligibility
- Community Edition — a standard self-hosted version of n8n available on GitHub
Data, GDPR & hosting
A consolidated view of how n8n handles your data.
GDPR overview
n8n GmbH is established in Berlin, so the GDPR applies directly and no Article 27 representative is required. The security page states plainly how the company complies with the GDPR, and an external Data Protection Officer is appointed, reachable at a dedicated address whose postal details are published. Data subject requests go to privacy@n8n.io, and the policy lists the rights under Articles 7(3), 15, 16, 17, 18, 20, 21 and 77. A pre-signed Data Processing Agreement can be downloaded and countersigned, a named list of sub-processors is published with each one's purpose and processing location, and transfers outside the EU rely on adequacy decisions, the EU-US Data Privacy Framework and standard contractual clauses. The policy in force is dated 24 June 2026 and earlier versions remain accessible.
Who owns the data?
Under the self-serve terms, everything you upload, submit or run through the cloud service — data, inputs, prompts, workflows and files — is Customer Content, and you keep all rights, title and interest in it. It stays private by default. You grant n8n a non-exclusive, royalty-free, worldwide licence, sublicensable to its subcontractors and service providers, solely to provide and support the service. Output produced by the AI features belongs to you as well. Two carve-outs deserve attention: workflows or posts you deliberately publish become Community Content, licensed broadly to n8n and to other users; and technical logs, metadata and telemetry, which exclude Customer Content, belong to n8n as Usage Data.
Reuse rights
You may use your own content freely: nothing in the terms restricts what you do with the data you bring in, the workflows you build or the output the AI features generate, and n8n claims no ownership over any of it. The licence you grant works one way only — it lets n8n process your content to run and support the service, not to exploit it. Beyond that, n8n commits in writing not to use Customer Content or Output to train any machine-learning model, and not to keep Customer Content longer than the AI feature needs to return a result. It may nonetheless analyse Usage Data and prompts internally to improve its AI services, in aggregated and de-identified form only, and may derive de-identified data sets from Customer Content to develop the platform. If you connect a third-party language model with your own API key, your data travels to that provider under that provider's terms, which you are responsible for reviewing.
Data retention & training
Hosting summary
For plans hosted by n8n, the platform runs on Microsoft Azure and both the physical hardware and the stored data sit in the European Union; the pricing FAQ specifies servers in Frankfurt, Germany. The published sub-processor list confirms Azure processing in Germany and Sweden and Hetzner Online in Germany, and names AI providers processing in the United States — OpenAI, Anthropic, Google Vertex AI, LangChain and Daytona — which come into play according to how you configure the platform. A group entity, n8n US Inc., is listed as an affiliate sub-processor in the United States. Cloud instances are logically isolated from one another and the database sits on a private network. Data at rest is encrypted with Azure Storage server-side encryption using AES256 in a FIPS-140-2 compliant implementation, backups included, and traffic in transit is protected with certificates managed through Cloudflare. If you self-host, hosting is wherever you put it and n8n has no control over it. The company states it is preparing additional hosting locations.
Things to keep in mind
Risks and trade-offs to weigh before adopting n8n.
- Use of the cloud service is limited to internal business purposes: reselling, sublicensing, leasing or transferring it is prohibited by the self-serve terms
- The AI features come with no accuracy guarantee — you remain solely responsible for validating output before deploying it, and high-risk uses are contractually excluded
- Annual commitments are non-refundable and there is no pro-rata refund; monthly plans only stop at the end of the current cycle
- Exceeding the Business quota is billed at 4,000 EUR per extra block of 300,000 executions, invoiced up to 45 days after the fact, and n8n may raise your tier automatically with 14 days' notice
- Subscribing to a Business plan does not cancel an existing subscription: you have to cancel the old one yourself
- On self-hosted instances telemetry is collected by default and must be switched off manually, and a Business or Enterprise licence key pings n8n's licence server daily to stay active
- Workflows shared on the community forum become Community Content, licensed broadly to n8n and to other users, and Customer Content is deleted six months after account deactivation, so export before you leave
Setup & Integrations
Technical difficulty
Moderate to demanding, depending on the route. The cloud version needs only an account and a free trial that asks for no credit card, and the first workflow can be assembled from one of more than 10,000 templates. Building anything real, though, means understanding APIs, credentials, webhooks and JSON. Self-hosting adds genuine operations work: Docker, a reverse proxy for TLS, disk encryption, upgrades and a licence key for paid features. Free courses, extensive documentation and an active forum soften the curve considerably.
Deployment
Integrations
Behind n8n
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement n8n.
Frequently asked questions
Can I run n8n on my own infrastructure?
How is n8n billed?
Is there a free trial, and is a credit card required?
Where is my data hosted?
Does n8n train AI models on my data?
Can I use my own language model?
Is there an API?
What support is included?
How long is my data kept, and can I get a Data Processing Agreement?
Should you pick n8n?
n8n is one of the better-documented tools in its category, and the paperwork backs the product up: a company registered in Berlin under HRB 212509 B, a full German imprint, eight distinct legal documents, four published price tiers, a documented REST API and a source repository with more than 200,000 stars. Very little here is claimed without being shown.
The positioning is unambiguous. This is a tool for technical teams, not a consumer assistant. The canvas is genuinely visual, but APIs, credentials, webhooks and JSON are unavoidable, and the payoff for that effort is a platform that does not stop where no-code tools usually do — a Code node in JavaScript or Python sits wherever you need it.
The structural decision a buyer faces is deployment. n8n's cloud keeps data in the European Union, on servers in Frankfurt, with a named sub-processor list and a security programme aligned to SOC 2. Self-hosting removes the residency question entirely, at the price of owning encryption at rest, TLS termination and upgrades yourself. Both routes come with the same written commitment: Customer Content and Output are not used to train machine-learning models.
The reservations are mostly commercial. The gap between the Pro plan and the Business plan is wide, dedicated support with an SLA only arrives at the Enterprise tier, and no permanent free plan appears in the cloud grid — the free route is self-hosting the Community Edition. Annual commitments are non-refundable and Business overages are billed in large blocks.
For an engineering, data or security team that wants automation it can inspect, extend and host itself, n8n is a serious candidate. For a team hoping to avoid technical work altogether, it is the wrong shape of tool.
- Choosing a selection results in a full page refresh.
- Opens in a new window.