n8n
Workflow automation for technical teams: a visual canvas with JavaScript and Python underneath, free self-hosted or cloud plans billed per workflow execution.
- Ratings
- No member reviews yet
- Price
- From $20/mo
- Last updated

Key takeaways
- n8n is n8n GmbH's workflow automation and AI orchestration platform, sold as n8n Cloud or self-hosted under a fair-code license. Reception is strong: G2 showed 4.7/5 from 287 reviews and Capterra 4.6/5 from 45 when checked in August 2026, and the GitHub repository displayed 202,793 stars the same month.
- It's built for technical teams, and that cuts both ways. The visual canvas gets a workflow started, then JavaScript or Python, the HTTP Request node and agent tooling take over; the dominant complaint across review sites is a steep learning curve for anyone who doesn't already know APIs and JSON.
- Pricing switched to euros: as of August 2026 Cloud Starter is €20 a month billed annually and Pro is €50, charged per complete workflow execution rather than per step. The self-hosted Community Edition stays free indefinitely.
- Third-party estimates put n8n.io at about 6.1 million visits in June 2026, down from roughly 9.2 million in November 2025. Over the same stretch the company's press-reported valuation climbed from about $2.5 billion at the October 2025 Series C to $5.2 billion after SAP's May 2026 investment.
- The biggest caveat is operational: self-hosting means Docker, SSL, scaling, backups and patching are yours, and August 2026 security advisories covering credential leakage and denial of service make version hygiene essential.
What is n8n?
n8n is a workflow automation and AI orchestration platform from Berlin's n8n GmbH. You connect applications, APIs, databases and language models on a visual canvas, then add branching, approvals and agent logic, and drop into JavaScript or Python when clicks run out. The homepage tagline as of August 2026 states the design plainly: "Code when you need it, UI when you don't." It runs as a hosted product, n8n Cloud, or as software you install on your own server.
Two things separate it from mainstream automation tools. The first is the license: n8n ships under a fair-code Sustainable Use License, so the self-hosted Community Edition is free indefinitely for internal business automation, though it isn't a conventional unrestricted open-source license. The second is billing: Cloud plans charge per complete workflow execution no matter how many steps it contains, instead of metering every operation the way per-task competitors do.
Since its 2019 founding the product has moved from integration plumbing toward AI orchestration: multi-step agents with tools, memory and human approval steps, plus an MCP server and client so Claude, ChatGPT or a coding agent can discover, build and run workflows. Investors followed the pivot, with a $180 million Series C in October 2025 and a strategic SAP investment in May 2026 at a reported $5.2 billion valuation.
How it works
Open a Cloud workspace or install your own
Create an account on n8n Cloud for a hosted workspace, or install the Community Edition on your own server. Starter and Pro trials need no credit card.
Build on the canvas
Add a trigger, drag nodes from the Node panel, connect branches visually and configure credentials for each service the workflow touches.
Add code, control and AI
Reach for expressions, JavaScript or Python Code nodes and the HTTP Request node where the point-and-click options end, then wire in LLM nodes, vector stores, tools and human-approval steps. Any compatible model provider connects with your own credentials.
Test, then operate
Run individual steps, replay or mock data, inspect inline inputs and outputs and evaluate AI behavior before activating. Once live, monitor executions and insights, expose workflows through the API or MCP, and export JSON or manage changes with Git where available.
Features
Visual workflow editor
A node canvas that combines drag-and-drop construction with inline data inspection, so you see what each step receives and returns as you build. Thousands of reusable workflow templates jump-start common builds; current repository material says 9,000 or more.
Code node
Write JavaScript or Python inside a workflow when configuring nodes visually won't cut it. It's the escape hatch that keeps complex transformations in the same canvas instead of in a separate service.
HTTP Request
Call arbitrary APIs, paste cURL-derived requests and build custom integrations for services without a dedicated node. Reviewers lean on it often, since incomplete integrations are a known gap.
AI Agent
Build multi-step agents with tools, logic, memory and approval paths, using the models and vector stores you choose. n8n is model-agnostic: its materials name connections to OpenAI, Anthropic, Google and open-source models, and you supply the provider credentials.
MCP Server and Client
Let Claude, ChatGPT or coding agents discover, create, edit and run workflows, and expose your own workflows as custom MCP tools. An August 2026 update added one-click OAuth connections to nearly 70 MCP servers.
Debugging and evaluation
Re-run a single step, replay or mock data, inspect logs and evaluate AI outputs without rerunning the entire workflow. Short feedback loops matter here because agent behavior needs checking before a workflow goes live.
Enterprise governance
Higher tiers document SSO/SAML/LDAP, RBAC, audit and log streaming, secret stores, environments and Git-based version control. An HTTP API and CLI support programmatic workflow, credential and execution management.
What people use it for
- Onboarding new employees and provisioning IT users.
- Enriching security-incident tickets and routing SOC work.
- Turning natural-language requests into API calls for DevOps.
- Summarizing customer reviews into sales and customer insights.
- Building chatbots and multi-step AI agents over company data.
- Scraping and summarizing webpages.
- Backing up workflows to GitHub and transforming JSON into Excel.
It's not the tool to hand a non-technical team as their first automation platform: reviews consistently describe a steep learning curve and documentation and support friction for users without prior API knowledge.
Who it’s for
Pricing
Official pricingIndividual plans
Community
Free
self-hosted
The full workflow engine on your own server, free indefinitely under the fair-code license.
Starter
€20
per month, billed annually
2,500 executions, 1 shared project, 5 concurrent executions, unlimited users, 2,300 AI Assistant credits; €24 billed monthly.
Pro
Popular€50
per month, billed annually
10,000 executions, 3 shared projects, 20 concurrent executions, 7 days of insights, admin roles, workflow history; €60 billed monthly.
Business plans
Business
€667
per month, billed annually
Self-hosted; 40,000 executions, 6 shared projects, SSO/SAML/LDAP, environments, Git version control, 30 days of insights.
Enterprise
Custom
contact sales
Custom executions, unlimited active workflows, 200+ concurrent executions, 365 days of insights, log streaming, dedicated SLA support.
As of August 2026 the published prices are in euros, a change from the dollar figures in older third-party captures, so confirm the currency at checkout. Cloud Starter is €20 a month billed annually or €24 monthly with 2,500 executions and 2,300 AI Assistant credits; Pro is €50 or €60 with 10,000 executions; the pricing material also documents a larger 50,000-execution Pro size at €120 a month. Business is €667 a month billed annually and self-hosted; Enterprise is contact sales. Billing counts complete workflow executions, not individual steps. AI Assistant credits are a separate monthly allowance that doesn't roll over and can't currently be topped up; the self-hosted AI Assistant uses your own keys with no credit limit. Starter and Pro trials need no card, while Business trials last 14 days and require one. Refund terms are narrow: monthly subscribers can request that month's payment back within 14 days of the billing date after cancelling, and annual plans run to term with no pro-rata refund. One unresolved discrepancy to note before you cancel: the self-serve terms say data is generally deleted six months after deactivation, while the Help Center says account data and backups are retained 90 days.
Traffic & reach
The numbers below are third-party estimates, a proxy for adoption rather than anything n8n publishes. They put n8n.io at about 6.1 million visits in June 2026, down from roughly 9.2 million the previous November. Read the slide alongside the rest of the record: the GitHub repository showed 202,793 stars in August 2026, review ratings held strong, and the company's reported valuation doubled to $5.2 billion in May 2026, so site traffic is one signal among several rather than the whole story.
Visits, June 2026
6.1M
▼ 9.5% month over month
Nov '25Jun '26
Top countries
- India15%
- United States14%
- Germany4.3%
- France3.8%
- Pakistan3.6%
Traffic sources
- Organic search38%
- Direct34%
- Paid search9.6%
- Referrals7.2%
- Organic social3.2%
- AI assistants3.1%
- Display ads2.5%
- Email1.6%
- Visits, last 3 months
- 19.6M
- Avg. visit
- 3m 00s
- Pages per visit
- 3.6
- Bounce rate
- 51.0%
- Brand searches / mo
- 21.4M
User reviews
Write a reviewNo reviews yet
Used n8n on something real? Write down what you made, what it cost and what the trade-off was. The next person can decide in about a minute from that.
Be the first to reviewWhat to know before you commit
What people rate it for. Reviewers consistently praise flexibility: visual workflows you can extend with code, custom APIs and complex branching, which makes n8n useful for quick automations and production pipelines alike. Self-hosting, broad integrations, AI features and cost efficiency for technical users recur as positives, and the aggregate ratings back it up, with G2 at 4.7/5 from 287 reviews and Capterra at 4.6/5 from 45 as of August 2026.
The learning curve is the complaint pattern. Non-developers struggle with JSON mapping, APIs, expressions and advanced nodes, and the documentation gets called sparse or inconsistent, so people end up in tutorials and forums instead. Trustpilot evidence runs more negative than G2: a June 2025 reviewer called the product good but said support and sales did not respond when the customer wanted to expand. No reliable current aggregate Trustpilot score was established.
Self-hosting is real operations work. Running n8n in production can require meaningful Docker, SSL, operations and scaling expertise, and reviewers report fragile workflows after major version updates. Integrations are sometimes incomplete. One reviewer built a separate Make.com scenario just to post to Pinterest because n8n has no integration for it, and others fall back on HTTP workarounds. Cloud pricing can also feel high for low-volume users, and cloud versus self-hosted feature parity is a recurring concern.
The 2026 security advisories make patching essential. In August 2026 n8n disclosed a credential-secrecy issue affecting the Strapi, SeaTable and Mailcheck nodes, patched in 1.123.73, 2.35.4 and 2.36.2, and a prototype-pollution issue in the Edit Fields node that could cause instance-wide denial of service, fixed in 1.123.67, 2.31.5 and 2.32.1. NVD also recorded CVE-2026-54305, which let an authenticated user enumerate or interfere with other users' dynamic credentials. March 2026 reporting described broader exploitation claims around n8n; the strongest narrative came from a Cloud Security Alliance research note and should be checked against n8n's own advisories rather than treated as a confirmed breach.
Fair-code, not open source. The self-hosted project uses n8n's fair-code Sustainable Use License, not a conventional unrestricted open-source license: internal business automation is permitted, but reselling or white-labeling n8n itself requires checking the current license and commercial terms. On data, the privacy policy says customer and integration personal data is not used to train n8n or third-party models, though de-identified datasets may be used to improve services and self-hosted telemetry has an opt-out.
The growth numbers are company or press reported. The valuations, about $2.5 billion at the October 2025 Series C and $5.2 billion after SAP's May 2026 investment, are company and press figures, not audited revenue evidence. Total funding raised is unreconciled because the SAP transaction was a secondary share sale with no disclosed amount, and the roughly 458-person headcount comes from a third-party record that is not independently confirmed.
Frequently asked questions
Company & links
- Company
- n8n
- Founded
- 2019
- Founders
- Jan Oberhauser
- Leadership
- Jan Oberhauser is the founder and CEO in company materials. A German developer and former freelance and agency technologist, he founded n8n in Berlin in 2019 after wanting a more flexible alternative to integration tools that charged per operation. A current full executive roster was not located.
- Funding
- Reported rounds: a $1.5 million seed in March 2020 co-led by Sequoia Capital and firstminute capital, a $12 million Series A in April 2021 led by Felicis Ventures, a €55 million Series B in March 2025 led by Highland Europe at a reported valuation of about €250 million, and a $180 million Series C in October 2025 led by Accel at about $2.5 billion. SAP made a strategic investment in May 2026 through a secondary share sale with an undisclosed amount, at a reported $5.2 billion valuation. Total raised is unreconciled across third-party databases.
- Scale
- A third-party organization record lists about 458 workers, which is not independently confirmed and should be treated as approximate. No reliable self-reported revenue, profitability or cash-flow figure was confirmed.
Try n8n on work you already know
A task you’ve done before is the fastest way to judge whether n8n fits your workflow and budget. Check the current plan terms on the official site before you subscribe.