Flurf —
your enterprise app,
now self-driving.

A real startup. A real brief. A unique design challenge.

Flurf is not a personal project — it is a commissioned design for a real product from Flurf Labs, a startup founded by a junior colleague based in Chennai. The brief was clear: design the browser extension UI for their AI-powered autonomous browser agent, and make it feel genuinely different from anything else in the market.

Flurf's product is technically ambitious. It sits as a browser extension inside enterprise applications and lets employees control the software using plain English — "create a new account for John" or "generate this month's sales report" — without clicking through menus. The AI handles the navigation. This was my first experience designing for an AI product, and it fundamentally changed how I think about interface design for intelligent systems.

Designing for an AI agent — not a normal app.

Before designing anything, I needed to deeply understand what Flurf actually does — because designing a UI for an AI agent is fundamentally different from designing a standard app. The interface isn't showing the user content. It's showing the user an intelligent system taking action on their behalf.

Dark, intelligent, minimal. Like the product itself.

The design direction was defined by one question: what does an intelligent system that takes action on your behalf look like? Not a toy. Not a standard SaaS tool. Something that feels genuinely powerful — like software that knows what it's doing.

The answer was a dark-first design system built around deep navy backgrounds, electric blue as the only accent colour, and an interface stripped of everything unnecessary. If a UI element doesn't help the user give a command or understand what the AI is doing, it shouldn't exist.

The decision to lead with dark theme was deliberate. Flurf's target users — enterprise employees navigating complex software — often work in dimly lit offices for extended periods. Dark mode reduces eye strain during long sessions while also making the blue action indicators and status signals more visually prominent. The light theme provides an alternative without compromising the visual hierarchy.

First contact with the product. Set the tone immediately.

The splash and login screens establish the visual identity of Flurf before any interaction happens. The splash uses the logo mark alone — three horizontal lines forming a stylised "F" shape — centred on pure dark navy. No text, no explanation. Just the mark. This builds recognition and communicates confidence.

The login screen uses dark card-style input fields on the navy background — keeping the enterprise feel consistent from the first screen. Google and Apple social login options reduce friction for enterprise users who are already authenticated through their organisation's SSO. The "New here? Click here" pattern at the bottom keeps sign-up accessible without cluttering the login flow.

The command centre. Minimal and immediately useful.

The home screen of the extension serves as the command centre. It opens with a search/command input at the top — the primary action users take — followed by a Quick Start grid of four pre-configured shortcut actions, and a History section showing recent conversations.

The most critical screen. Where commands become actions.

The chat interface is the heart of Flurf — where users give commands and watch the AI execute them. Designing this screen required solving a unique problem: how do you show an AI agent taking action in real time in a way that feels reassuring rather than mysterious?

The answer was the Action Status Card — a four-stage indicator that shows exactly what the AI is doing at each moment in the execution pipeline.

The 4-stage action status card — shown inline in the chat after every command execution

Every command. Instantly replayable.

One of Flurf's core features is Command History — the ability to access and re-execute any previous command. The history screen shows all past conversations in a sortable list. This is particularly powerful for enterprise users who run the same complex workflows repeatedly — a "generate monthly report" command used every month can be replayed in one tap.

Why each decision was made.

Every visual and structural choice in the Flurf extension UI was made in response to the client brief, the product's technical context, or the specific needs of enterprise users.

What designing for a real startup taught me.

Flurf was my first client brief — designing for someone else's product vision rather than my own. It was a completely different experience from personal projects, and it taught me things I couldn't have learnt any other way.

• Designing for AI is a new discipline. A standard app shows content. An AI agent shows intent, action and state. The Action Status Card — Received, Processing, Action Initiate, Completed — emerged from trying to solve that fundamental difference. Showing what the AI is doing at every step was the most important UX decision in the entire project.
• Client briefs force you to solve someone else's problem. With personal projects you naturally gravitate toward problems you understand. With Flurf I had to deeply understand an enterprise AI product I had no prior experience with. That research — reading the website, understanding the competitive landscape, understanding why enterprises resist new tools — directly shaped every design decision.
• The dark-first direction was a risk that paid off. Most browser extensions use white interfaces. The brief asked for something unique. Going dark-first was the boldest interpretation of that brief — and it created a design identity that immediately sets Flurf apart from every other extension in the Chrome store.
• Delivering both light and dark themes is harder than it looks. It's not simply inverting colours. Every component has to be reconsidered — shadows that work on light backgrounds don't work on dark ones, border weights that define cards on white need to be lighter on dark, accent colours need different opacity levels depending on the surface they sit on. This project significantly improved my understanding of design systems.
• If I were to continue, I'd design the onboarding flow. The current screens cover the core extension experience. A complete v2 would include a guided first-use experience — connecting Flurf to a specific enterprise application, setting up the AI training, and designing the first successful command moment. That onboarding is where most enterprise software adoption fails, and getting it right is critical for Flurf's growth.