Every morning I open the same tabs.
Figma. Notion. Analytics. Docs.
One by one. Every. Single. Day.
Small frustration — but small frustrations compound. So I decided to fix it.
What I Built
Toolflow is a Chrome extension that lets you:
- Group your tools into named workflows
- Open everything in one click
- Launch a full Chrome tab group in seconds
- Drag to reorder, rename, edit — everything
Here's a 90-second demo:
👉 Get Toolflow — $4.99 lifetime
The Honest Story: I Wrote Zero Code
I directed Claude (Opus 4.6) to build the entire thing. Every file — the manifest, background service worker, popup UI, CSS, all of it — was written by Claude.
What I actually did:
- Came in with one clear, specific idea
- Made every product and UX decision myself
- Described not just what I wanted but why
- Reviewed everything and redirected when something felt off
My job was clarity. Claude's job was implementation.
How It's Structured Under the Hood
For those curious about the architecture — it's a standard Manifest V3 extension with three core files:
background.js — A service worker that handles URL meta-fetching. When you paste a URL into the Add Tool modal, it fetches the page's <head> and extracts the OG title and meta description automatically. It reads only the first 50KB to avoid large downloads, uses an 8-second abort timeout, and falls back gracefully if the fetch fails.
popup.js — All state management, rendering, drag-to-reorder logic, CRUD operations, and navigation live here. No framework — pure vanilla JS with Chrome's storage API for persistence.
popup.css — A milky light theme with liquid glass UI components. The buttons use backdrop-filter: blur() with layered border gradients to achieve the glass effect without any external library.
The whole thing is intentionally zero-dependency. No React, no build step, no bundler. Just HTML, CSS, and JS — which is exactly what a Chrome extension needs to be fast and lightweight.
What Directing AI Actually Looks Like
This is the part most people get wrong when they say “AI built it.”
AI doesn't make product decisions. It implements them.
When I told Claude I wanted an auto-fetch feature for URLs, I didn't just say “add auto-fetch.” I described:
- When it should trigger (on blur and on paste)
- What it should fill in (name and description only if fields are empty)
- How it should handle failure (silently, no error shown to user)
- What the loading state should look like (spinner, then checkmark, then idle)
That level of specificity is what separates a good AI-built product from slop. The taste is mine. The speed is Claude's.
What Kept It Fast
- One clear idea before starting — I didn't scope-creep mid-build
- No design tool — straight to code, faster feedback loop
- Shipped rough first — polished after it worked, not before
Total time: 2 days to build, 1 day to record the demo, build the landing page, and launch.
What This Means for Developers
The bottleneck in building products used to be implementation speed.
That bottleneck is largely gone for people who learn to direct AI well.
The new bottleneck is clarity. Can you describe what you want precisely enough? Can you spot what's wrong with something you didn't write yourself? Can you make fast decisions about what to cut?
Those are human skills. They're worth more now, not less.
Try It
If you open the same tabs every day, Toolflow was built for you.
🔗 toolflow.moyrith.com — $4.99, lifetime access
Follow me on X @Moyrith where I build in public.
Would love to hear what workflows you'd set up — drop them in the comments.
Top comments (2)
This is a great example of where clarity matters more than coding.
I tried something similar for a small internal tool, and the hardest part wasn’t building — it was deciding exactly how it should behave. AI wrote the code fast, but the real work was refining the UX and removing friction.
Also love that you kept it zero-dependency. Chrome extensions feel much faster and easier to maintain that way.
100% the UX decisions were way harder than the build itself. You can prompt your way out of bad code, you can't prompt your way out of a vague idea.
Zero-dependency was a no-brainer for an extension.