If I started a company today, I would spend my first €20 on a server, not €400 on software subscriptions before a single feature shipped.
Ten years ago, founding a startup came with a fixed entry fee. Slack for chat, Figma for design, Mailchimp for the newsletter, Zapier to glue it all together, Zoom for calls. Maybe €400 a month before a single feature shipped, just to look like a company that knew what it was doing.
That entry fee has quietly collapsed, and most founders haven't updated their mental model yet. I've spent the last months writing about workshop tools, vibe coding my own apps, and testing self-hosted alternatives on a cheap rented server. It changed how I think about startup costs entirely.
The honest answer today is simple: you can run a company on open source infrastructure for the price of a server and a couple of hours setting it up. Open source infrastructure means software whose source code can be inspected, modified, and run on infrastructure you control. For many business categories, that is the better default.
The bubble that actually matters isn't the AI one
I wrote a while back that the AI investment bubble will probably burst the way the dot-com one did, while the underlying capability quietly stays and reshapes everything anyway. Something similar is happening one layer down, in ordinary business software.
For a decade, venture-funded SaaS could subsidize a beautiful product with cheap capital and win on polish alone. That capital is getting more expensive, prices are creeping up, and meanwhile a parallel universe of lean, self-hosted, often single-maintainer tools has quietly caught up on the fundamentals.
Penpot does 90 percent of what most teams use Figma for. Plane covers what most teams actually need from Jira. Cal.com does everything Calendly does and provides the source code. None of these are compromises anymore. They're what I'd reach for first.
What I'd actually install, in the order a company needs it
If I were setting up a company from zero tomorrow, the first thing I'd buy isn't software. It's a €20-a-month server, and I'd put Coolify on it, a self-hostable deployment interface that turns one server into the base for almost everything below.
Getting the team working together
Before there's a product, there's a group of people who need to plan, decide, and not lose track of who's doing what. I'd kick off with a proper kickoff workshop, planned in Sessionplan. Yes, it's my own tool, and I'd use it precisely because it needs no account and stores nothing you'd have to explain to a lawyer later.
Project tracking runs on Plane instead of Linear, mostly because I'd rather pay for a bigger server than per-seat pricing that punishes us for growing the team. Shared credentials live in Vaultwarden from day one, not in a spreadsheet somewhere.
Building the thing
Design work goes on Penpot, self-hosted, so the whole team, including whoever we hire from outside the design bubble, can comment and prototype without a seat licence.
Internal docs and specs live in AppFlowy or a plain wiki on Outline, whichever the team actually opens without complaining. The tool matters less than having one place where decisions survive the meeting that produced them.
Talking to the outside world
Once there's something to show, scheduling and client or investor calls run through Cal.com. The newsletter runs on Listmonk, and Plausible tracks the website without triggering a cookie banner. That alone saves a legal conversation.
Automation between all of it runs on n8n rather than Zapier, partly for cost, mostly because I don't want our workflows locked behind someone else's pricing tier the moment we exceed 100 tasks a month.
What I wouldn't self-host: payments, email delivery, and probably the core hosting for the actual product if it needs to scale fast. Some things are genuinely worth paying a specialist for, and pretending otherwise just moves the risk from your subscription bill to your own server's uptime.
Stripe stays Stripe. The point was never to self-host everything on principle. It's to stop paying rent in categories where a mature, boring, well-maintained open source project already does the job.
This is the same shift I keep writing about
I built a gym tracker last year because I was tired of paying for an app I barely used, and shipped it to the Play Store in three months without being a "real" developer in the traditional sense. That wasn't a fluke. It was the same democratization of production I keep coming back to on this blog: production tools that used to require capital or a specialist team are now available to anyone willing to spend a weekend reading docs.
Open source infrastructure is the business side of that same story. The cost of looking like a serious, well-run company has collapsed, and most founders haven't noticed yet because the SaaS sales pages are still louder than the GitHub repos.
That's also, not coincidentally, why I keep the tools I build myself, like Sessionplan, free and open. If the barrier to building software is disappearing, the barrier to giving it away should disappear right alongside it.
What this means in practice
Spend your first money on the product and on talking to actual users, not on the software stack around it. Rent one server, put Coolify on it, and pick three or four tools from the open source side before you reach for a credit card.
You'll spend a Saturday afternoon on setup instead of €400 a month, and you'll still look, to any outside observer, exactly like a company that knows what it's doing. Because you will be one.
Read next:
- The democratization of production: why production tools are moving out of specialist hands
- Vibe Coding for Designers: The Real Risk Nobody Talks About: the same shift applied to software creation
- Workshop meeting planner: a practical example of building a focused tool yourself
FAQ
Is self-hosted open source software reliable enough for a company to depend on?
For the categories mentioned here, yes. Penpot, Plane, Cal.com, and n8n are used in production by teams far larger than a typical startup. The risk shifts from a vendor raising prices to keeping the server patched, which is a trade many technical founders can manage.
Do I still need someone technical to set all of this up?
Less than you might think. Coolify handles much of the deployment complexity with a Heroku-style interface, and tools like Sessionplan or Plausible's hosted option need no server at all. A founder comfortable with a terminal can have the core stack running in an afternoon.
What should I never self-host, even as a technical founder?
Payments and transactional email delivery, at minimum. Both carry compliance and deliverability requirements that specialists solve better than a side project ever will. Core product hosting should also move to a proper provider once you need to scale quickly.
Does this work for agencies and non-technical teams too?
It works for both, with a different starting point. Non-technical teams should begin with tools that need zero setup, such as Sessionplan, Plausible's hosted option, or Cal.com's cloud plan. They can add self-hosting later once there is budget or technical help.
Is self-hosting only about saving money?
No. Cost is the visible part, but the bigger shift is control. Your data stays on infrastructure you own, you are less exposed to a vendor's next price increase, and you can adapt the tool instead of filing a feature request and waiting.