You know the feeling. A new idea hits you. Maybe it's a SaaS you want to test, an internal tool for a client, or a side project you're genuinely excited about. You open the terminal, scaffold a project, and start building.
Then reality kicks in.
Authentication. You need authentication. An hour later, you have OAuth and sessions working. Payments. Stripe integration, webhook handling. There goes a day. Email templates. Background jobs. Analytics. Error tracking. Workspaces with team invites.
And you haven't written a single line of product code.
I've done this dance many, many times. Each time I told myself "this time I'll be faster." Each time I wasn't. The same infrastructure, the same edge cases, the same debugging sessions.
So I built Eden Stack, a production-ready starter kit for AI-native SaaS applications. Not as a business plan. As a solution to my own recurring frustration.
Here's the honest story of how it happened and what I learned.
The Pattern That Kept Repeating
I'm 26 years old, working as a developer and consultant in Oslo. Since ChatGPT launched in late 2022, I've built multiple AI-native applications, some for clients, some for myself.
Two problems kept appearing:
Problem 1: Infrastructure setup took 60+ hours before writing product code. Auth, payments, workspaces, email, analytics, error tracking, blog, SEO. Important, necessary, but not my product. Every project needed the same foundation, just configured slightly differently.
Problem 2: Users now expect AI features as standard. Chatbot, intelligent search, document processing, background AI tasks that deliver results even when you close the browser. Building these from scratch takes weeks. For each project.
The creative energy for a new idea has a half-life. Every hour spent on infrastructure setup is an hour where your motivation decays. By the time the foundation was ready, the excitement for the actual product had often faded.
After the fifth time, the thought was obvious: what if I build a foundation designed to be shaped? Where you can swap pieces in and out based on what your specific product needs?
What I Actually Built
Eden Stack is a monorepo with everything a modern SaaS needs, pre-wired and working:
The SaaS basics (all edge cases handled):
- Authentication with OAuth providers
- Multi-tenant workspaces and organizations
- User settings and team invites
- Stripe payments with webhook handling that never silently fails
- Email templates
- Background jobs with automatic retries
- Analytics and error tracking
AI features (production-ready, not prototypes):
- Agentic chatbot where the AI decides when to search the web, process documents, or just respond
- Web search and deep research that runs for minutes and returns reports with citations
- Document processing: upload a PDF, get it chunked, embedded, and ready for AI chat
- Long-running AI tasks that survive if the user closes the browser or the server restarts
AI-powered development:
- 30+ Claude skills that encode the exact patterns of the codebase
- MCP setup that configures your entire project (database, payments, email) in one prompt
- Everything designed so AI follows YOUR architecture, not generic patterns
You buy it once, you own the code, you adapt it to your needs.
The Ownership Philosophy
This is the part I care most about, and it's inspired by shadcn/ui.
Before shadcn, UI components meant installing a library locked to its constraints. Or building everything from scratch. shadcn changed the game: beautifully designed components that live in your codebase, not in node_modules. No wrapper components to fight. No breaking changes from library updates.
Eden Stack applies the same thinking to the full stack.
Need Azure PostgreSQL instead of Neon? Describe the change to Claude, and it migrates the code. Prefer AWS Step Functions over Inngest? Claude helps you swap it. Each integration is isolated in its own module. Removing Stripe doesn't break authentication. Swapping the database doesn't affect API routes.
You're not locked in. You're starting from a working system and making it yours.
The LinkedIn Post That Surprised Me
When I first posted about Eden Stack on LinkedIn, I expected my developer friends to be interested. What surprised me was the breadth: developers, founders, consultants, people in large companies exploring agentic development.
The common thread wasn't "I want a boilerplate." It was: "I know what I need to build. I just don't want to spend weeks getting to the starting line."
That validated something I'd suspected. The problem isn't a lack of knowledge or tutorials. The problem is the gap between "I know what I need" and "it works in production," and the weeks it takes to cross that gap.
Pricing: Why $99
I went back and forth on pricing. Free felt wrong. I wanted people who'd actually use it and give feedback, not collectors. Too expensive and I'd limit the audience.
I landed on $99 early bird, $199 regular. One-time purchase. You own the code.
The math: the AI features alone take 2-3 weeks to build properly. The SaaS infrastructure is another week-plus. At $99, you're paying less than one day of freelancer rates for a month of work.
More importantly, the price is low enough that buying it is an easy decision, but high enough that buyers are serious about actually building something.
Four Things I've Learned
1. Users expect AI features. It's table stakes now.
When I started building, "AI-native" felt like a differentiator. By the time I launched, it was table stakes. New SaaS products without AI features feel incomplete. This isn't hype. It's what users tell me.
The specific features that resonate: chatbot with tool calling, document processing (upload a PDF, ask questions about it), and background AI tasks that work even when you're not watching.
2. "Edge cases handled" resonates more than feature lists
When I describe Eden Stack, the thing that makes people's eyes light up isn't "we have auth and payments." Every boilerplate has that. It's "all the edge cases are handled."
What does that mean concretely? Stripe webhooks that use durable execution so they never silently fail. Auth that works across web and mobile with the same session. Organization invites that handle every state (pending, accepted, expired, revoked). Background jobs that automatically retry on failure.
These aren't features. They're the boring, critical work that separates a demo from a product.
3. Two types of "AI-native," and you need both
"AI-native" has two dimensions:
Dimension 1: AI features for your users. Chatbot, search, document processing, background AI tasks.
Dimension 2: AI-powered development. Claude skills that encode your codebase's patterns, so AI doesn't guess. It follows your architecture.
Both matter. The first makes your product competitive. The second makes you fast at building it.
4. The agentic development mindset is a real skill
I increasingly think of AI as a team of junior developers I'm directing. Give each one a clear task, clear constraints, and let them work in parallel. This isn't a metaphor. I literally use Git worktrees where different AI agents work on different features simultaneously.
But it requires clear product vision per agent, tasks that are self-contained enough to run in parallel, and quality assurance. The orchestration is a skill you build over time. I believe it's the skill that will separate developers who thrive in the AI era from those who feel threatened by it.
What’s Next
I'm building in public, which means being honest about what works and what doesn't. Right now:
- The web experience is solid
- The AI features are production-tested
- The development workflow with Claude skills genuinely changes how fast I ship
If this resonates with you, if you've felt the frustration of setting up the same infrastructure for the third or fifth time, or if you're curious about what an AI-native starter kit actually includes, I'd love your feedback. That's worth more than a sale.
The tools have never been better. The threshold has never been lower. You don't need a team of ten. You need a good idea, the right tools, and the willingness to ship.
Just build.