Most companies don't realise how much they are paying for software that they could get for free. Not cheaper. As a gift. The same function, the same level of maturity, the same tidy interface, just without the item that comes out of the account every month and gets a little bigger with every new employee, every gigabyte of memory and every API call.
The catch isn't in the price. It's in the word gift: the licence is free, the operation is not. If you don't separate the two, you're making one of two mistakes. Either they pay SaaS subscriptions for years for functions that have long been freely available as mature open source software. Or they underestimate the labour costs of their own server and end up with a subsequent problem. Both errors stem from the same gap: Not doing the maths. It is precisely this calculation that has shifted in 2026, and the trigger is called Coolify.
Why the make-or-buy question is now new on the table
Two events have brought the self-hosting discussion out of the tinkering niche in 2026.
Firstly, Heroku, for years the epitome of "just deploy, someone else will take care of the server", switched to so-called sustaining engineering mode in February 2026. This means no more new features, only bug fixes and security updates. A platform on which many had built half their product business was de facto frozen.
Secondly, the cost debate surrounding platforms such as Vercel has finally gone viral. The best-known case is the creative platform Cara, whose number of users shot up from around 40,000 to around 650,000 accounts in one week and which received a Vercel bill of around 96,000 US dollars overnight. An extreme case, for sure. But it has revealed a pattern that many people are familiar with on a smaller scale: With usage-based cloud pricing, success is a financial risk. Just when things are going well, the bill explodes.
Coolify has filled this gap. In April 2026, after almost two years in beta, version 4.0 was released. In spring 2026, the project had over 50,000 stars on GitHub, making it the most widely recognised self-hosted PaaS ever. This is no hype coincidence. It hits a nerve.
What Coolify is, and what it isn't
Coolify is an open source platform that turns your own server into something that feels like Heroku or Vercel. The difference becomes clear when you put the before and after side by side.
Before: renting a VPS, writing Docker compose, configuring a reverse proxy, applying for SSL certificates and renewing them on a regular basis, maintaining deployment scripts. With Coolify: connect a server via SSH, link a Git repository, click on "Deploy". The rest is done by the layer in between. It builds the Docker image, starts the container, automatically retrieves a Let's Encrypt certificate, terminates TLS via a reverse proxy and deploys the app under your desired domain. Git push, deployment is running. That's the promise that made Heroku famous in 2010, only on hardware that belongs to you.
The real leverage lies in the more than 280 preconfigured services. These are ready-made templates for databases, caches, analytics, CMS, backend stacks and around two dozen developer tools. Instead of laboriously setting up software yourself, you select it from a list and Coolify rolls it out as a container. It is precisely this library that turns "theoretically you could host it yourself" into "it will be up and running in fifteen minutes".
In short: Coolify is not the software that you end up using. It's the layer that makes running third-party open source software so easy that the make-or-buy question needs to be revisited. And self-hosting Coolify is permanently free: Apache 2.0 licence, no feature gate, no user cap, no trial period. There is a paid cloud version starting at around $5 a month for the administration level, but you don't need it to get started. Your only ongoing expense is the server that runs everything, and a decent VPS starts at around $5 a month.
Keep one thought in mind for the rest of the article: Coolify lowers the barrier, it doesn't remove it. Once the services are up and running, you are the operator, the SRE of your own little platform. We'll explain what this means in concrete terms below. First, the good news about what you can do with it.
The best SaaS alternatives self-hosted: These tools you run right now
For almost every SaaS category that SMEs pay a subscription for today, there is a mature open source alternative that you run on your own infrastructure via Coolify. I sort them not as a uniform list, but according to what they are in everyday working life: what your visitors and readers see, what works invisibly in the background and what holds your most sensitive data.
What is visible to the outside world
Those who use Google Analytics today do not pay with money, but with data and with legal effort. Plausible and Umami are both lean, data protection-friendly analysis tools without personalised tracking, which makes the cookie banner superfluous in many constellations. Plausible has over 20,000 GitHub stars, a useful indicator of serious distribution. Both run as a Coolify template. You get reach and conversion figures without passing on your visitors' data to an advertising company in the USA. For a German company, this is an argument of trust towards its own customers, not just a data protection footnote.
On the publishing side, there is Ghost. It reduces blogging to the essentials: fast pages, distraction-free editor, in-built SEO, memberships and subscriptions. If you use Substack or Medium, you're building your audience on someone else's turf, with all the rules and commissions dictated by the provider. Ghost self-hosted gives you the same convenience, but the content, reader list and revenue remain yours.
What works in the background
Here's the one tool where self-hosting almost always pays off: n8n, the open-source answer to Zapier and Make. It's a visual automation platform with over 400 integrations, and the reason for the clear recommendation is the pricing model of the competition. With Zapier and Make, you pay per task or per operation. If you take automation seriously, you quickly run into four-digit annual amounts because every single workflow run is billed. The self-hosted Community Edition of n8n is fair-code-licensed and free of charge, including unlimited executions. You pay for the machine, not the runs. That's also why n8n is a separate topic here and forms the backbone of many automation and AI projects.
Right next to it is Supabase, an open-source backend platform consisting of a Postgres database, authentication, automated APIs, file storage and real-time functions. It is the most serious self-hostable alternative to Google's Firebase and is relevant for anyone building an app or product. It's the difference between "we're chained to a cloud provider" and "our backend runs where we want". And for structured data that would otherwise end up in Airtable, there's NocoDB: an Airtable-style table interface whose data resides in your own Postgres or MySQL database. NocoDB runs as a single Docker container and, like n8n, is available under a fair-code or source-available licence, meaning it can be operated freely, but is not a classic open source in the OSI sense.
What holds your most sensitive data
Nextcloud is the closest thing to a self-hosted Google Workspace: file synchronisation and sharing, plus calendar, contacts, mail, a collaborative office suite and even encrypted video calls in Teams or Zoom style with Nextcloud Talk, all on your server. For companies that work with sensitive data or don't want to entrust their internal communication to a US provider, this is the central pillar of digital sovereignty. To be honest, Nextcloud is the heaviest tool on this list and needs to be maintained.
Vaultwarden is an extremely resource-efficient implementation of the Bitwarden API written in Rust. You use the official Bitwarden clients, but the vault is on your server: Access data, secure notes, files, everything locally, with no cloud dependency and no subscription fee per user. And finally, Twenty is a modern, API-first open source CRM with a customisable data model, a serious entry point for smaller sales teams that don't need the full HubSpot or Salesforce apparatus but don't want to let their customer data out of their hands.
For those who want to see the categories side by side:
| Category | Open source tool (self-hosted) | Typical SaaS alternative |
|---|---|---|
| Web-Analysis | Plausible / Umami | Google Analytics |
| Publishing / Newsletter | Ghost | Substack, Medium |
| Process automation | n8n | Zapier, Make |
| Backend / App Platform | Supabase | Firebase |
| Structured Data | NocoDB | Airtable |
| Files & Collaboration | Nextcloud | Google Workspace |
| Password Management | Vaultwarden | 1Password, Bitwarden Cloud |
| CRM | Twenty | HubSpot, Salesforce |
This is what a realistic entry looks like
So that this doesn't remain abstract, here is the path from zero to the first running service, without embellishment. First you need a server. A VPS with two CPU cores and four gigabytes of RAM from a European provider is enough to get you started; Hetzner and similar hosters charge in the low double-digit euro range per month. If you are planning several heavyweight services such as Nextcloud and Supabase in parallel right from the start, you should rather go for eight gigabytes, otherwise things will get tight.
The Coolify installation itself is a single command that you execute as root on the fresh server. You then access the interface in the browser, create an admin account and connect the server on which Coolify is running as the first deployment target. From here, it's just a matter of clicking: In the template area, select Plausible, for example, enter the desired subdomain to which a DNS entry points and start the deployment. Coolify pulls the image, starts the containers, retrieves the SSL certificate and reports an accessible, encrypted URL after a short time.
There are two things you should not put off until later, even if the temptation is great. Firstly, backups: Coolify can write automated database backups to S3-compatible storage, and this is exactly what you set up before the first real data set is in the system, not after. Secondly, access: a separate user instead of root for everyday use, a restrictively configured SSH access and a firewall that only leaves the necessary ports open. This is not rocket science, but it is the borderline between "runs securely" and "runs until someone finds it". If you work cleanly here, you will have a basis in just a few hours on which further services only cost minutes.
What "free" really costs
Now the part that most lists on this topic omit. Self-hosting is not free, it's just paid for differently. It's exactly the steps you just took, backups, updates, hardening, that are the core of what is now yours. The licence costs nothing, the server costs a few euros, but you take over the operation. You are now the SRE. If the VPS provider has a network fault, the Coolify interface is unavailable until it's fixed. When Docker eats up storage space, you're the one who cleans it up. Updates, backups, monitoring, restarting after an outage - it's all part of the job, and it's exactly the work you're actually paying the SaaS provider for.
That's why the honest answer to "is it worth it?" is a calculation, not an ideology, and one that takes your own working time into account. As a rough guide, not as a law of nature: below around 50 to 100 euros in cloud costs per month, the switch is rarely worthwhile because the operating costs eat up the savings. Above a few hundred euros per month, in many cases it tips in favour of self-hosting, with savings that, depending on the load, can be noticeably more than half and, in favourable cases, approach 70 percent. The footnote behind this is important: This margin only applies if you realistically offset the ongoing in-house maintenance and security costs. If you don't do this, every self-hosting bill looks too good.
This is precisely where Coolify becomes the factor that shifts the calculation. It reduces operating costs to such an extent that the threshold at which self-hosting becomes worthwhile is lowered. What used to require an administrator now runs via an interface that a technically skilled employee can operate on the side. This doesn't change the basic truth that someone has to take care of it, but it makes that "someone" much smaller.
When-you-should-do-it
I'm CTO enough to say this clearly: there are cases in which self-hosting is the wrong decision. If your team has no Docker or Linux experience and doesn't want to build any, you're buying yourself a follow-up problem with the supposedly free tool. If the application is business-critical and every minute of downtime costs money, you need an operational concept with readiness and redundancy, not a single VPS with good intentions. And for highly fluctuating, hard-to-predict loads, the elastic scaling of the big clouds is a real advantage that a fixed server can't replicate.
Self-hosting is not a statement of faith against the cloud. It is a tool that is clearly the better choice in certain constellations and not in others. The mistake is not to use SaaS. The mistake is never doing the maths.
What this means for medium-sized companies
For a medium-sized company, the exciting part is not the individual savings on the password manager. It's the sum of three things: predictable cost control instead of usage-based invoices that grow with success; data sovereignty, because customer data and internal communication are stored on infrastructure under German and European jurisdiction, which makes data protection and compliance much easier; and independence, i.e. no platform risk, no frozen Heroku, no surprise bill as with Cara.
A simple model calculation shows how big the leverage can be. Take a medium-sized company with 40 employees that currently has five SaaS tools on subscription: Analytics, password manager, file sharing, a small CRM and an automation platform. That quickly adds up to several hundred euros a month, often with per capita prices that increase with every hire. The same functions can be bundled on one or two VPSs via Coolify, with basic costs in the low double-digit to low triple-digit range. The difference is real, but it's not the whole gain. The real gain is that the question is no longer "if", but "who will run it".
And this is where the decision is rarely "completely self-run" versus "completely rented". The pragmatic approach is usually a setup that someone sets up, secures and documents once, and that your team then operates on a day-to-day basis. We therefore don't think of self-hosting as a DIY project, but as an architectural decision with an operating concept: which services belong on your own infrastructure, what backups and monitoring look like, and how the whole thing fits into your existing individualised software landscape. Where automation and AI processes come into play, self-hosted n8n is the natural hub anyway, a topic that runs directly into our AI consulting.
The sober conclusion
The provocative thesis from the beginning that you pay for software that you could get for free is true with a footnote: the licence is a gift, not the operation. What has really changed in 2026 is not that open source software has suddenly become good. It has been for years. What has changed is that tools such as Coolify have lowered the operating hurdle to such an extent that the old standard assumption that "renting is always easier" no longer automatically applies.
The right reaction is therefore not a change of belief, but a table. What does SaaS cost us today, what would in-house operation cost, including the honestly calculated working time, and where does data sovereignty count so much that it makes the difference? Anyone who does this calculation properly will make the make-or-buy decision with their eyes open for the first time. And the break-even point is more often on the side of your own server than the old standard assumption would suggest.
Lay your five most expensive SaaS subscriptions on the table. We will honestly calculate the case, tell you for each tool whether it is worth running it yourself and, if necessary, set up the operating concept once, from the individual software team to n8n automation. Talk to us.