I replaced 14 tabs with one chat window

I vibe-coded my own dashboard and it's ugly and perfect

Daniel Plomp

04 Mar 2026 — 5 min read

I used to open 14 tabs every morning. Stripe for revenue. Gmail for support tickets. Analytics for traffic. My bank for cash flow. Notion for tasks. Each tool had its own login, its own dashboard, its own way of showing me numbers I needed to cross-reference manually.

Last month I replaced all of that with a single chat window.

"Show me yesterday's revenue, failed payments, and customer bug reports."

Done. One message, one answer, thirty seconds. I didn't touch a single dashboard.

I didn't switch to some fancy new product. I connected my existing tools to an AI agent through MCPs (Model Context Protocol servers — basically APIs that agents can use). The agent reads from Stripe, my database, Gmail, everything. I just ask for what I need.

And that broke something in my brain about how software works.

You use 5% of every tool you pay for

Think about the last SaaS product you signed up for. How many features does it have? Hundreds, probably. How many do you actually use? Five. Maybe ten.

I went through my entire stack and it's embarrassing. Stripe has maybe 200 features. I use the dashboard and the payment links. My email tool has automation builders, template galleries, A/B testing. I use "send email."

I was paying for 100% of every product and using 5%. We all are.

The reason I chose most of these tools was their interface. On the database level, they all do roughly the same thing. Stripe and its competitors store the same payment data. Email tools send the same emails. Project management tools track the same tasks. The UI was the differentiator — the layer between me and the data that made one tool feel better than another.

That layer just became optional.

The dashboard is dead

When my agent can read and write to any tool through an API, the UI stops mattering. I don't need a pretty chart when I can ask "what's my MRR trend this quarter?" and get a direct answer. I don't need an onboarding flow when the agent already knows the API.

I ended up doing something I didn't expect: I vibe-coded my own dashboard. Just the 5% I actually use from each tool, all in one place. Took a weekend. It's ugly and it's perfect.

This is the part where traditional SaaS founders should get uncomfortable.

Your UX moat? Gone. Your onboarding funnel? Nobody goes through it. The agent doesn't care about your landing page. It cares about your API.

The race to zero

Once the interface doesn't matter, what's left? The API. And APIs are commodities.

I don't care if my email goes through Resend or Postmark or SES. I don't care if my database is Supabase or PlanetScale or Neon. It just needs to work and cost the least.

I'd actually love it if my agent handled this automatically. Every day, check if there's a cheaper alternative. If there is, migrate my data and cancel the old one. No switching costs. No migration headaches. Just the cheapest option, always.

The logical endpoint: I wouldn't even know which tools I'm using. Their brands wouldn't stick in my mind because I never see their interfaces. They become invisible infrastructure.

That's a nightmare if you're building a SaaS with 95% margins. When every tool is interchangeable and an agent handles switching, margins go to zero. You become a utility.

So what still works?

I've been thinking about this constantly. If traditional SaaS is toast, where do you build?

Network data that agents can't replicate. Stripe's fraud detection is trained on billions of transactions across millions of businesses. My agent can switch me to a cheaper payment processor in five minutes. But it can't bring that collective intelligence with it. If your product genuinely gets better because more people use it, you have something agents can't commoditize.

Boring regulated industries. My agent auto-switches my email provider without asking. Would I let it auto-switch my bank? My insurance? No. Licenses, compliance, and trust are moats that have nothing to do with code quality. Not sexy, but durable.

Selling results instead of access. I don't want to pay $49/month for a social media tool anymore. My agent does the three things I used it for. But I would happily pay someone to get me 50 qualified leads this month from LinkedIn. That's not a tool, it's a result. And I don't care how they do it behind the scenes. The pricing flips from "pay for access" to "pay for outcomes." You can charge more than SaaS ever could, because you're competing on "did it work?" and results have clear ROI.

The agent layer itself. I built my own agent setup because I'm a technical founder who's been shipping for years. Most people can't do that. Someone needs to build the orchestration layer that collapses 30 tools into one chat for normal people. That's a real product. Probably the biggest one in this whole shift.

One-time code, not subscriptions. For most of my side projects, I'd happily pay someone $199 for a repo I can throw into Claude Code and customize. Start from a solid default instead of from zero. No recurring fee, just a good starting point. Sell 10,000 copies at $199 and you're doing $2M a month. When everyone stops paying monthly for SaaS, paying once for code becomes an easy decision.

This has happened before

Every time something dies in tech, the thing that replaces it is bigger.

Websites killed newspapers but created Google. Mobile killed the desktop web but created the app economy.

AI agents are killing SaaS. What they're creating doesn't have a name yet. I don't know exactly what it looks like. But I know the winners won't be the ones with the best dashboard. They'll be the ones who figured out what to build on top of the agent layer.

The uncomfortable question

If you're building a CRUD app with a nice UI and calling it a SaaS business, this should keep you up at night. Not because you're wrong today — the revenue is real, the customers are real. But because the ground is shifting under you and the pace is accelerating.

The question isn't whether agents will replace your UI. They will. The question is whether you have something underneath that UI that agents can't replicate or route around.

If the answer is just "a database with a pretty frontend" — you already know what happens next. You just don't want to think about it yet.

In this series: