$0 Cloud Bills: Indie Hackers Gut PaaS Incumbents
How a self-hosted Heroku alternative and Oracle’s free ARM instances are reshaping what’s possible for bootstrapped builders
The first time I shipped a full LLM-powered SaaS with zero infra cost, I laughed out loud. Then I checked the bill again. Still zero.
- Indie hackers are deploying production AI SaaS apps with $0 monthly infrastructure cost using Coolify, Oracle’s free ARM instances, and Cloudflare.
- Bootstrapped founders and solo builders benefit; PaaS vendors like Vercel, Render, and Netlify lose leverage as lock-in weakens.
- This mirrors the 2015 Dokku/Flynn wave but succeeds due to ARM-friendly AI tooling and sustained free-tier infrastructure from cloud providers.
- Watch for Oracle tightening instance policies or Coolify adding managed hosting—either could break or validate the trend.
“The first time I shipped a full LLM-powered SaaS with zero infra cost, I laughed out loud,” said a founder in a Slack thread last week. “Then I checked the bill again. Still zero.”
He wasn’t joking. No trial. No promo code. No ticking clock.
Just a side project,an AI tool for summarizing legal discovery docs,running on a 4-core, 24GB RAM ARM server, hosted by Oracle, accessible via HTTPS, backed up to S3, and deployed through Git. All free. Forever.
The stack? Coolify, self-hosted on Oracle’s always-free tier, fronted by Cloudflare Tunnel. No Vercel. No Netlify. No AWS bill at the end of the month.
And it’s not just him. Across indie hacker forums, GitHub repos, and weekend hack threads, this setup is quietly becoming the default for launching small AI products. Not because it’s flashy. Because it works,and costs nothing.
The Deployment
Coolify is an open-source, self-hostable alternative to Vercel, Heroku, and Railway. It lets developers deploy websites, APIs, databases, and services to any server they control,whether it’s a VPS, a Raspberry Pi, or a corporate EC2 instance,using Docker and Git-based workflows.
It’s not a new project. But its relevance has surged in 2026, thanks to two external shifts: Oracle’s sustained commitment to its free-tier ARM instances and the rise of lightweight AI SaaS built by solo founders.
Oracle offers a 4-core, 24GB RAM ARM64 instance,machine size: A1.Flex,for free, indefinitely. No credit card required. No time limit. You can run it forever. The catch? It’s ARM, not x86. But with Docker support and growing LLM tooling for ARM (like Ollama, Llama.cpp, and lightweight inference servers), that’s no longer a blocker.
Coolify runs on that instance. Once installed, it becomes a control panel for deploying apps. Push code to GitHub, and Coolify pulls it, builds the container, handles Let’s Encrypt SSL, and routes traffic. You can deploy a Next.js frontend, a FastAPI backend, a PostgreSQL DB, and a Redis cache,all from one UI, all on one box.
It supports one-click deployments for over 280 services. Want MinIO for object storage? Click. RabbitMQ? Click. PocketBase for a no-code backend? Click. All run in isolated containers, managed by Coolify’s lightweight orchestration layer.
And because it’s self-hosted, you’re not locked in. All configuration lives on your server. If Coolify shuts down tomorrow, your apps keep running. You just lose the dashboard.
The stack, as it’s now circulating in indie circles, looks like this:
- Oracle Free Tier: 4-core ARM, 24GB RAM, 200GB storage
- Coolify: Self-hosted on the Oracle box, manages deployments
- Cloudflare Tunnel: Exposes services securely without opening ports
- Cloudflare Pages: Serves static frontends (Next.js, Astro, etc.)
- PocketBase or Supabase: Lightweight backend, deployed via Coolify
- S3-compatible backup: Backups go to Oracle Object Storage or another S3 endpoint
Total monthly cost: $0.
[[IMG: a solo developer in a Lisbon co-working space reviewing deployment logs on a laptop, Cloudflare and Coolify tabs open, coffee cooling beside the keyboard]]
Why It Matters
This isn’t just about saving money. It’s about changing who gets to build.
For years, the indie hacker playbook relied on platform-as-a-service tools,Vercel, Netlify, Render,that made deployment easy but came with trade-offs: cost at scale, vendor lock-in, limited control, and opaque pricing.
Vercel’s Pro plan is $20/month. Fine for a portfolio site. But if you’re shipping an AI app that spikes in usage, costs balloon. One user’s LLM-heavy workflow can trigger $300 in egress fees. Suddenly, your “free” tier isn’t free.
Coolify flips that. You pay nothing to Oracle. You deploy what you want. You control the server. You set the limits.
And because Coolify is open source and self-hosted, there’s no corporate roadmap to derail your stack. No sudden pricing changes. No deprecation notices.
It’s particularly powerful for AI indie hackers,those shipping small, focused tools powered by open-weight models (Mistral, Llama 3, Phi-3). These aren’t billion-parameter behemoths needing GPU clusters. They’re lightweight, fine-tuned models running on CPU or integrated into efficient inference pipelines.
A tool that processes PDFs, summarizes text, or extracts entities doesn’t need Kubernetes. It needs a server, a database, and a way to deploy code. That’s all Coolify provides.
Compare this to 2015, when Dokku and Flynn tried to bring Heroku-like ease to self-hosting. They failed to gain traction because cloud platforms made deployment too easy. Why manage a server when you could git push heroku master?
But today, the calculus has changed. Platform lock-in is riskier. Costs are less predictable. And developers,especially those building AI tools,want control.
Coolify isn’t the only player. Ploi, Caddy, and CapRover offer similar self-hosted PaaS experiences. But Coolify’s UI is cleaner, its one-click services broader, and its CLI tight enough to integrate with AI assistants for natural-language infra debugging.
This is the quiet return of the generalist founder: someone who writes code, deploys infrastructure, and ships products,without a DevOps team.
And Oracle? They’re not doing this out of charity. They’re betting that today’s indie hackers are tomorrow’s enterprise customers. Let them build on ARM now. When they scale, they’ll need more compute, more storage, more services,and Oracle’s full cloud suite will be waiting.
What Other Businesses Can Learn
If you’re a small team or solo founder shipping a lightweight AI product, this stack is worth testing. Not for every use case. But for side projects, MVPs, or niche tools with moderate traffic, it’s viable.
Here’s how to approach it:
1. Start with the Oracle Free Tier.
Sign up for Oracle Cloud. Choose the “Always Free” ARM instance. Pick the A1.Flex shape: 4 cores, 24GB RAM. Install Ubuntu 22.04 or 24.04. That’s your base.
2. Install Coolify.
The CLI install is one command: curl -s https://get.coolify.io | bash. It guides you through setup, including domain and SSL. You’ll need a domain (Cloudflare makes this easy). Once installed, access the UI at yourdomain.com.
3. Deploy your first app.
Link a GitHub repo. Choose a framework,Next.js, Express, Flask. Coolify auto-detects the build process. It creates a Docker container, exposes it on a subdomain, and sets up HTTPS.
4. Add a backend.
Use the one-click PocketBase option. It gives you a NoSQL database, auth, and file storage. Connect it to your frontend via API.
5. Secure and expose.
Don’t open ports. Use Cloudflare Tunnel (cloudflared) to expose your Coolify dashboard and apps. This way, only Cloudflare can reach your server. No public IP exposure.
6. Back up.
Set up automatic S3-compatible backups. Oracle Object Storage works. So does Backblaze B2. Coolify handles the rest.
7. Monitor.
Use Coolify’s built-in monitoring for CPU, memory, disk. Set up Telegram or Discord notifications for crashes or high load.
“You’re not saving money,you’re removing a constraint.”
That’s the real shift. When infra cost drops to zero, you stop optimizing for pennies and start shipping for speed.
Yes, there are trade-offs. ARM means some x86-only images won’t run. You’ll need to rebuild or find alternatives. No managed PostgreSQL? You manage it yourself. No auto-scaling across regions? You’re on one box.
But for a solo founder or two-person team, that’s often fine. You’re not building the next Slack. You’re building a tool that solves a narrow problem,for lawyers, therapists, teachers, or small manufacturers.
And if it grows? You can migrate. Move to a paid VPS. Add load balancing. Use Kubernetes. But you’re not forced into it prematurely.
One developer in Bristol told me they’d been using Render for a customer support chatbot. Cost: $89/month. After moving to Coolify + Oracle, they cut it to $0. “The only thing I miss,” they said, “is the ‘I don’t have to think about it’ feeling. But now I want to think about it. I like knowing where my code lives.”
[[IMG: a small-team engineering lead in a Denver basement office reviewing server resource usage on a second monitor, Coolify dashboard showing container metrics, whiteboard with migration checklist in background]]
Looking Ahead
Last week, I watched a founder in a Berlin livestream deploy a full AI SaaS in 17 minutes. Frontend. Backend. Database. Auth. HTTPS. All from a coffee shop.
No credit card. No invoice. No investor.
Just code, a domain, and a free server.
That’s the quiet revolution. Not AGI. Not $5B models. But the return of the self-sufficient builder,someone who owns their stack, their data, and their destiny.
Coolify isn’t perfect. It’s not for every team. But for a growing cohort of indie hackers, it’s enough.
And Oracle’s free tier? It’s not going away. They’ve renewed it three years running. The bet is clear: give early, win later.
The rest of us just get to watch,$0 at a time.
- Coolify, accessed 2026-04-26
- Oracle Cloud Free Tier Documentation, accessed 2026-04-26
- Cloudflare Tunnel Setup Guide, accessed 2026-04-26
More from the same beat.
5 Devs, 1 tmux Session: Agent of Empires Guts AI Workflow Chaos
A lightweight session manager for AI coding agents is quietly solving a real workflow pain point—one tmux session at a time.
- Agent of Empires is a Rust-based session orchestrator that manages multiple AI coding agents in persistent, isolated tmux sessions with optional Docker sandboxing and mobile-accessible monitoring.
Claude Code v2.1.118 Tightens Grip on Agent Runtime, Crowds Out UI-First Tools
A point release with signal: better vim ergonomics, tighter plugin control, and the quiet arrival of MCP tooling at the hook level.
- Claude Code v2.1.118 shifts from chat-based coding assistant to an operational runtime for autonomous agents, with vim visual modes, unified usage tracking, and MCP tooling at the hook level.
500 EU Firms Ditch US Clouds as Cohere Absorbs Aleph Alpha
A Canadian-German tie-up backed by Lidl's owner is pitching itself as the non-American option. Read the contract before the press release.
- Cohere is absorbing Aleph Alpha with backing from Schwarz Group and government support, creating a unified sovereign AI stack for EU and Canadian mid-market enterprises unwilling to rely on America…