n8n Review 2026: Is Self-Hosting Worth It for Founders?
n8n review posts usually land in one of two camps. There is the breathless AI-agent hype, and there is the developer tutorial that assumes you enjoy patching servers for fun. Neither camp tells you what the thing actually costs or who should bother with it. That is the gap this review tries to fill, because parts of my own operation run on n8n alongside Make, and the two tools keep teaching me different lessons about cost and control.
Something else changed the conversation this year. In May 2026, SAP put money into the company behind n8n at a $5.2 billion valuation, a strange kind of milestone for a tool you can still self-host for free. More on that in a minute, because it cuts both ways for small teams.
What n8n Actually Is
n8n is a workflow automation tool you build the same way you wire up a diagram. Nodes connect in the order data flows through them: one node pulls an email, the next asks a model to summarize it, a third posts the summary where your team will see it. Drag, connect, test. At a glance it looks like Zapier.
Two things set it apart. First, the escape hatch: any workflow can drop into a Code node where you write plain JavaScript or Python when the prebuilt steps fall short. Second, the whole thing can run on your own hardware, free, without asking anyone’s permission.
A few facts worth knowing up front:
- It is fair-code, not open source in the formal sense. The Sustainable Use License lets you run it freely for your own business, but you cannot repackage it and resell it as your own hosted product.
- The GitHub repository crossed 200,000 stars in August 2026, and n8n itself lists 400+ integrations.
- Workflows can branch and loop. Failed runs can retry, and error paths route messy cases somewhere visible, which matters once a workflow grows past the demo stage.
- A cloud version exists where the maintenance is not your problem at all.
Crowd sentiment matches reality here. On G2, n8n’s seller page shows 4.7 out of 5 across roughly 290 reviews, with a heavy tilt toward five stars. Reviewers mostly praise the flexibility and grumble about the learning curve, which lines up with what you will read in community forums.
Why the SAP Investment Matters (and to Whom)
On May 12, 2026, n8n announced a strategic investment from SAP valuing the company at $5.2 billion, more than double its worth from under a year earlier. The announcement included a partnership: n8n technology will be embedded inside SAP’s Joule Studio on the SAP Business AI Platform. Two takeaways. The first is that SAP did not buy n8n outright, it invested and locked in a distribution path into enterprise accounts. The second is that free self-hosted edition is still shipping, and community workflows keep working the way they always have.
For a solo founder this is mostly good news with one asterisk. The good news is resources: the company can keep developing both tracks (cloud and self-hosted) while serving enterprise demand. The asterisk is that big-company needs now weigh more on the roadmap than they did before. That last part is my read rather than a documented fact, but it is the usual pattern when a scrappy tool gains a corporate partner.
How n8n’s Pricing Works
n8n bills per execution, and its own pricing FAQ defines that plainly: one execution is one entire run of a workflow, no matter how many steps it contains. Loop over 500 rows and send 500 Slack messages inside a single workflow? That is one execution. A workflow that fires every 15 minutes? That is around 2,900 executions a month, because the trigger is the counter, not the work inside.
Compare that to the metered alternatives. Make charges a credit per module execution, and Zapier bills per task. A twelve-step pipeline that runs once costs one execution on n8n, roughly twelve credits on Make, and twelve tasks on Zapier. If your workflows are short and shallow, the difference barely registers. If they are long and run at volume, n8n quietly becomes the cheapest option in the room.
There is a wrinkle nobody puts in the comparison charts. Deep workflows favor n8n’s math, while shallow ones that fire constantly rack up executions just as fast as they would rack up credits on a per-step platform. The pricing page has a monthly and annual toggle, and annual billing saves about 17%. Paying month to month works out to roughly €24 on Starter instead of €20, so the annual commitment is real money saved if you plan to stay.
Current n8n pricing lives on its pricing page, which is also the source of truth, since it has already changed once this year.
n8n Plans at a Glance
| Plan | Price | Executions per month | Runs on | Notes |
|---|---|---|---|---|
| Community Edition | Free | Unlimited (your hardware) | Self-hosted | You manage updates and security |
| Starter | €20/mo billed annually | 2,500 | n8n cloud | Unlimited users and workflows |
| Pro | €50/mo billed annually | 10,000 | n8n cloud | Longer history, admin roles |
| Business | €667/mo billed annually | 40,000 | Self-hosted license | SSO, Git version control, queue mode |
| Enterprise | Custom | Custom | Cloud or self-hosted | SLA support, external secret stores |
Two details in that table deserve their own sentences. Every paid plan includes unlimited users and unlimited workflows, with every integration available, which removes the seat-counting headache that most subscription pricing creates. And n8n advertises a 50% discount on the Business plan for qualifying startups with fewer than 20 employees, which turns €667 into something an early-stage team could genuinely swallow.
Now the warning. On cloud plans, executions over your monthly limit do not queue and wait. They fail on the spot per n8n’s own support documentation, and there is no automatic catch-up when the month resets. A lead-intake workflow that jams on the 25th stays jammed until you upgrade. Set a usage alert before this happens. It’s the kind of detail that separates a review from a brochure, so expect me to keep repeating it.
Self-Hosting: Free Software, Real Responsibilities
The Community Edition costs nothing to license. Spin up a small VPS, run n8n in Docker, and you’re off, with no execution cap because the only limit is your server. People commonly run small personal instances on virtual servers from roughly $5 to $50 a month depending on load, though treat that as an estimate rather than a promise.
My own situation is modest. I keep a small n8n instance running in Docker on a Linux box alongside other containers, and it handles a handful of steady jobs: content checks, data pulls, the plumbing this site’s operation leans on. Maintenance is manageable now, but I will say this honestly. The first month wasn’t zero-effort. Upgrades and backups are on me, along with the occasional restart, and n8n’s documentation says the same thing in drier language: self-hosting requires technical knowledge for setup and configuration, for securing servers, and for scaling work. They recommend the cloud version if that isn’t your thing.
There is also the license question. The Sustainable Use License is free and generous for running your own operation, yet it isn’t open source, and it doesn’t permit offering n8n to others as a competing service. Most founders will never trip on this. If you’re building a product where n8n is essentially the product, read the license first.
What I Like (and Where It Bites)
- Per-run billing rewards deep workflows instead of punishing them.
- Code nodes give you an escape hatch when a connector cannot do the thing you need, with no waiting on a vendor’s roadmap.
- Execution history lets you open a failed run, see which node broke, and replay it after you fix the problem.
- Self-hosting is a real free tier, not a crippled trial.
The honest downsides. n8n has a genuine learning curve, one of the most common complaints on G2 and in forums, and I felt it myself in the first weeks. Cloud plans fail executions abruptly at the cap, as covered above. Self-hosted setups inherit uptime duty along with updates and backups as personal chores. None of these are dealbreakers. They are the tax you pay for a tool that does not hold your hand or meter your every step.
The AI Side: Agents and MCP
This is where n8n has moved fastest. In late September 2026, the team introduced AI Agents as a first-class feature: you describe what an agent should do, pick a model, choose the tools and workflows it may call, and let it run inside your automations without hand-wiring the whole chain yourself. Around the same time, one-click connections to dozens of MCP servers arrived in the node panel, which matters if you are wiring agents into a wider stack.
Keep the cost math honest in your head, though. Paid cloud plans include a monthly batch of AI credits, but the model calls themselves run through whatever provider credentials you connect, and that provider bills separately. An agent-heavy setup can quietly become a second line item on another vendor’s invoice. That isn’t an n8n quirk so much as a cost you have to track yourself, and it rarely shows up in marketing pages.
n8n vs Make vs Zapier for Founders
These three get compared constantly, and honestly, the right answer depends less on features than on who maintains the thing. I wrote a full Make.com review recently if you want that deep dive, so here I will keep the contrast tight.
| n8n | Make | Zapier | |
|---|---|---|---|
| Billing unit | One execution per workflow run | Credits per module (trigger is free) | Tasks per action |
| Entry price | €20/mo annual for 2,500 runs | $9/mo Core for 10,000 credits | Paid tiers start a bit higher |
| Self-host | Yes, free Community Edition | No | No |
| Deep workflows | Same cost regardless of steps | Each step burns credits | Each step burns tasks |
| Learning curve | Steepest of the three | Manageable for non-developers | Slickest onboarding |
| Best for | Technical founder with deep pipelines | Visual thinkers who like maps | Fastest first automation |
Make’s pricing, verified against its own plans in late September 2026, holds Free at 1,000 credits with a 15-minute minimum interval, Core at $9 for 10,000 credits, Pro at $16, and Teams at $29 for role management and shared templates. If a non-technical teammate needs to maintain your automations without learning a developer’s tool, Make is the comfortable choice. Its current plans are easy to model against the number of scenarios you actually run.
Zapier gets a different kind of honesty from me. I haven’t paid for Zapier myself, so I won’t fake a hands-on verdict about their daily interface. What I can say with confidence comes from what the company publishes and what reviewers agree on: the most polished user experience in the category, the widest list of native app connectors, plus pricing built around that convenience. Exact tiers sit on Zapier’s pricing page, and the task math matters more than the sticker price when your zaps grow.
One caveat from the broader comparison holds here too. Switching tools mid-flight is possible but rarely free, since rebuilding a mature scenario map by hand takes a weekend you would rather spend on your product, and export formats do not translate cleanly between platforms. It is recoverable, just not effortless. For the longer three-way breakdown written for bootstrappers, see Make vs Zapier vs n8n from earlier this year.
How I Ran This Evaluation
Full transparency, since trust is the only currency a one-person site has. I run automation daily as the backbone of my own setup, and n8n plus Make are both paid parts of that stack. Pricing and plan limits, along with license terms, were pulled from vendor pages and fresh third-party trackers in the first week of October 2026, with the SAP valuation and the agent launch cross-checked against primary announcements. Review-site figures come from n8n’s G2 seller page rather than aggregated blog claims. I am a solo founder building this kind of stack in public, as the About page explains, so my lens throughout is what works without a team behind you.
Key Takeaways
- n8n charges per workflow run, not per step, so deep multi-step pipelines are where it wins on price.
- Cloud starts at €20/month (annual billing) for 2,500 executions, and self-hosting the Community Edition remains free.
- Over-limit cloud executions fail immediately instead of queueing, so set a usage alert from day one.
- AI agents and MCP connections are current features, but your model provider bills those calls separately.
- If a non-technical teammate must maintain automations, Make or Zapier will cause less friction.
Frequently Asked Questions
Is self-hosting n8n really free?
Yes. The Community Edition carries no license fee and no execution cap; you pay only for the hardware it runs on. The catches are operational, since updates become your job, along with backups and security work. The license also bars you from selling n8n itself as a hosted service, which only matters if the automation is your product.
What happens when I hit the 2,500 execution cap on Starter?
Per n8n’s support docs, executions that arrive after the cap fail on the spot rather than queueing. Nothing replays automatically when the counter resets, so the practical move is watching usage through the month and upgrading before you hit a wall.
Do I need to know how to code?
No, though it helps. The visual editor handles ordinary automations without writing anything, and the Code node exists for moments when a connector falls short of what you need. If you plan to hand maintenance off to someone non-technical, weigh Make or Zapier first, since their interfaces are friendlier to that handoff.
Where to Start
If n8n sounds like your kind of tool, try the cloud Starter plan on the monthly toggle before committing to annual billing, or spin the Community Edition up in Docker on a machine you already pay for. And if your bigger problem right now is not automation but demand, my piece on getting your first 100 customers is the natural companion read. Build the plumbing, then give it customers worth serving.
Disclosure: some links in this article are affiliate links. If you buy through one, the site earns a commission at no extra cost to you, and I only recommend tools I would run myself.
