Why I Chose a Chrome Extension Over a Web App
When I decided to build a tool to track AI token usage, the first decision was the hardest: web app or Chrome extension?
Most developers default to a web app. It is what we know how to build. It is easier to deploy, update, and monetize. Browser extensions feel like a niche, difficult format.
I chose the extension. Here is the thinking behind that decision and whether it was right.
The core question: where does the user have the problem?
The problem I was solving was getting cut off by Claude or ChatGPT mid-session with no warning. The user has this problem inside the AI platform — not somewhere else.
A web app would require the user to:
- Notice they are getting close to a limit
- Open a new tab
- Navigate to my web app
- Check their usage
- Go back to their AI session
That is five steps to solve a problem that needs a zero-step solution. The information has to be visible in the same context where the problem occurs. That is only possible with an extension.
This reasoning is not specific to my product. It applies to any tool that solves a problem occurring inside a third-party website. If your product needs to know what the user is doing on another website, a Chrome extension is almost always the right format.
The distribution advantage nobody talks about
Web apps require you to build all your own distribution. SEO, paid ads, social media, content marketing — all of it is your responsibility.
Chrome extensions have a built-in distribution channel: the Chrome Web Store. Users search for "Claude token tracker" directly in the Web Store and find you. This is a different quality of intent than someone who finds you through a blog post — they are actively looking for a tool to solve a specific problem.
In my first 100 users, approximately 30 came from Chrome Web Store organic search. I did nothing to drive these installs. They happened because the extension existed and users searched for the category.
A web app would have zero equivalent of this. The Chrome Web Store is a search engine for tools, and being in it is a distribution asset.
The trust advantage
Chrome extensions require explicit user permission to install. This creates a higher-intent user than a web app where someone can land without any real intent.
A user who goes through the process of finding your extension, clicking "Add to Chrome," reviewing the permissions, and confirming installation is more invested than someone who stumbled onto your website. In my experience, extension users have significantly higher retention than web app users for equivalent products because the installation itself signals intent.
The technical constraints that actually help
MV3 Chrome extensions have strict limitations:
- No remote code execution
- No persistent background processes
- Strict Content Security Policy
- Limited API surface
These sound like disadvantages. They are actually product advantages in disguise.
No remote code execution means no external scripts, no tracking pixels, no third-party code running in the user's browser. This is a genuine privacy feature that users care about. "All data stays local" is not marketing copy — it is a technical constraint of the format.
No persistent background processes means the extension cannot quietly run in the background consuming resources. Users trust extensions more when they know they are not running 24/7.
The constraints forced TokenPulse to be a genuinely privacy-respecting, lightweight tool. I could not have built a tracking-heavy, analytics-heavy version even if I wanted to.
The real disadvantages
Updates go through Chrome Web Store review. When I fix a bug, I cannot push the fix instantly. I submit an update and wait 1-3 days for review. For a web app, fixes are live in seconds.
Discoverability outside the Chrome Web Store is harder. You cannot rank in Google for "TokenPulse" the same way a web app can. Extensions need a separate marketing site for Google visibility.
Monetization is less flexible. Chrome extensions cannot use Stripe directly in the extension UI without a backend. Subscription billing requires a web component. The Chrome Web Store takes a 30% cut on in-app purchases.
Cross-browser support is work. Chrome, Edge, Brave, and Arc use the same extension format and one build works for all. Firefox uses a different API. Safari is different again. Supporting all browsers requires ongoing work.
Was it the right choice?
Yes, unambiguously for this specific product.
The tool needs to be visible inside AI platforms. That requires an extension. There is no web app equivalent that would be as useful.
The Chrome Web Store distribution was a significant benefit — roughly 30% of early installs came from organic store search with no effort on my part.
The privacy constraints forced better product decisions than I would have made otherwise.
For a different product — one that solves a problem outside a third-party website, or one where users need cross-device access to their data — a web app might be the right choice. The format should follow the problem location.
The question to ask: where does the user have the problem? Build there.
TokenPulse — free Chrome extension for tracking AI usage.
