The problem
On X (Twitter), engagement means visibility. The more you reply, the more people see your profile. But writing smart, relevant replies in the right tone to dozens of tweets a day is a full-time job. Content creators and startup founders waste hours on manual engagement.
Reply Pilot automates exactly this: it generates contextual AI replies directly on X, without sounding generic or spammy.
What we built
Three components: the Chrome extension (the core product), the marketing website with landing page, and an API proxy on Cloudflare Workers. Plus LemonSqueezy integration for licenses and subscriptions.

The Chrome extension
A Manifest V3 extension that integrates natively into the X interface. We inject a button on every tweet: click it and you get Generate (quick generation) and Advanced (fine-tuned filters). The generated reply is inserted directly into the X composer, ready to post or edit.
- 4 tones: Professional, Casual, Witty, Friendly
- 3 lengths: Short (5-10 words), Medium (under 180 chars), Long (under 280 chars)
- Custom rules and style instructions per user
- Image understanding: analyzes tweet photos via Claude
- Quoted tweet and thread detection for full context
- One-click regeneration if you don't like the reply
- Learns from edits: the extension remembers how you modify replies and adapts its style
- Works on x.com, twitter.com, and X Pro (TweetDeck)


Settings page injected into X
Reply Pilot adds an item to the X sidebar that opens a full dashboard directly inside the platform. From here, users see their stats (generated vs posted replies, time saved, conversion rate), set style rules, enter their license key, and configure AI behavior. All without leaving X.

How the AI works
The extension extracts the full tweet context: text, author, images, quoted tweets, threads. Everything is sent to a Cloudflare Worker that forwards the request to Claude Haiku 4.5. The API key is secured server-side, never exposed in the extension. Rate limiting: 60 requests/minute per IP.
The learning system remembers the user's last 5 edits and includes them in future prompts, adapting the style to the user's actual personality.
Landing page
Marketing site built on Next.js 16 with React 19, dark design with gold accents, Framer Motion animations, and conversion-optimized sections: Hero, Features, How It Works (interactive carousel), Testimonials, Pricing (3 plans with monthly/yearly toggle), FAQ. Analytics through Pirsch.io (privacy-focused, no cookies).
Licenses and subscriptions
Monetization through LemonSqueezy: 3 plans (Starter $7/month, Pro $15/month, Unlimited $31/month with yearly billing). The user receives a license key on purchase, enters it in the extension, and can generate replies according to their plan. Monthly usage tracking stored locally in Chrome Storage, offline license validation with a 7-day grace period.
Tech stack
Chrome extension: JavaScript ES6, Manifest V3, Chrome Storage API, DOM manipulation for X integration.
Landing page: Next.js 16, React 19, TypeScript, Tailwind CSS, Framer Motion, Pirsch.io analytics.
Backend: Cloudflare Workers (API proxy with rate limiting), Claude Haiku 4.5 (AI), LemonSqueezy (payments and licenses).
Results
Product published on Chrome Web Store, conversion-optimized landing page with transparent pricing. Serverless architecture (Cloudflare Workers) keeps operating costs near zero. The edit-learning system makes replies better with every use.


