LaunchKit - AI Launch Copy Generator

Overview

LaunchKit is an AI-powered web application that turns a GitHub URL or plain-text project description into a complete launch package — instantly, with no signup required.

Paste your repo URL or describe what you built, and LaunchKit generates: a full landing page (headline, subheadline, feature bullets, CTA), platform-tuned social posts for Twitter/X, LinkedIn, and Product Hunt, and a single-sentence perfect pitch for bios and cold emails. Everything streams in real time as the model generates.

Live at launchkit.kroszborg.co

Problem Statement

Launching a side project requires writing the same information in five different formats for five different audiences — a landing page, a Twitter thread, a LinkedIn post, a Product Hunt tagline, and a one-liner pitch. This context-switching writing work takes hours and often gets skipped entirely, leading to projects that ship quietly with no audience. There was no tool that handled all of it from a single input.

Key Features

Landing Page Copy Generator

• Generates a complete landing page copy structure from your project description
• Headline, subheadline, three feature bullets, and a call-to-action — all in one pass
• Output is launch-ready with no editing required for most projects
• Informed by a 26KB landing page template (lib/landing-template.ts) engineered for high-converting structure

Platform-Specific Social Posts

• Twitter/X – Concise, punchy post optimized for the 280-character constraint and developer audience tone
• LinkedIn – Professional narrative post with structured formatting for broader reach and discoverability
• Product Hunt – Tagline and description formatted specifically for PH launch day community norms
• Each platform gets its own dedicated prompt with channel-specific tone, length, and structural rules

Perfect Pitch Generator

• One-sentence product description engineered for bios, cold emails, and intro conversations
• Captures what the product does, who it's for, and why it matters — no filler language
• Structured to work as a standalone sentence in any context

No Signup Required

• Fully functional without any account, payment, or API key
• GitHub URL support: fetches the repository's raw README, strips Markdown, uses it as project context
• All 4 output types stream in real time with a typewriter animation as the model generates
• Results appear section by section — no waiting for everything to finish before seeing output

Tech Stack

• Next.js 15.3.0 – App Router with Turbopack for fast development and edge-ready deployments
• React 19.1.0 – Server components and streaming-compatible component tree
• TypeScript 5.8.3 – Strict mode throughout with full type coverage
• Google AI SDK (@ai-sdk/google) – Gemini model integration
• Vercel AI SDK (ai ^6.0) – streamText, streamObject, and useStreamableValue patterns
• Tailwind CSS 4.2.1 – Latest v4 architecture with no config file
• Bun 1.3.8 – Package manager and runtime for fast installs and dev server
• Zod – Schema validation for structured AI output parsing

Technical Implementation

AI Streaming Architecture

LaunchKit uses the Vercel AI SDK's streamText for open-ended prose generation and streamObject for structured output sections. Each output type (landing page, Twitter, LinkedIn, Product Hunt, pitch) runs as an independent streamed request so sections appear progressively as they complete. The frontend renders each stream via useStreamableValue with a custom typewriter animation that displays characters as they arrive, giving users immediate signal that generation is working.

Prompt Engineering

The generation pipeline uses five separate, purpose-built prompts — one per output type. Each prompt specifies exact tone, length constraints, structural requirements, and persona for the target platform. A large landing page template (lib/landing-template.ts, 26KB) serves as the structural blueprint for the landing page generator, anchoring the model to proven high-converting copy patterns. Prompts are tested against edge cases: tools with no README, highly technical projects, and consumer-facing apps.

GitHub README Parsing

When a GitHub URL is provided, LaunchKit fetches the repository's raw README via the GitHub raw content CDN (raw.githubusercontent.com), strips Markdown formatting (headings, bold, links, code blocks) to plain text, and injects the cleaned content as project context into the generation prompts. This allows accurate copy for existing projects without requiring any manual description — the README does the work.

Design System

LaunchKit uses a deliberate editorial aesthetic designed to feel like a premium developer tool:

• Background: #0a0a0a near-black with no gradients
• Accents: warm off-white #e8e0d0 and muted ochre
• Typography: Playfair Display (serif headlines), DM Mono (body and code output)
• Components: flat outlined buttons, dark inputs, sharp-cornered cards
• Motion: fade-up reveals on load, streaming typewriter text, no bounce or spring animations

Architecture

app/
  page.tsx          — Hero with product explanation and input form
  generate/         — Generation interface and streaming display
  api/              — Backend routes for AI calls and GitHub fetching
lib/
  ai.ts             — AI integration utilities and prompt builders
  landing-template.ts  — 26KB structural template for landing page generation
components/
  ui/               — Design system components

What I Learned

Building LaunchKit taught me:

• How to compose multiple independent AI streams that each update separate UI sections simultaneously
• Vercel AI SDK's streamObject for getting structured JSON output streamed token-by-token
• Next.js 15 App Router edge runtime patterns for low-latency AI endpoint routing
• How prompt structure — not just prompt content — determines output consistency across runs
• Bun as a drop-in runtime for Next.js: faster installs, compatible dev server, no config changes

Impact

LaunchKit removes the most time-consuming part of launching a side project — writing the copy — and makes it a 30-second task.

• Zero friction – No account, no payment, no configuration required
• 4 outputs simultaneously – One input generates landing page, 3 social posts, and a pitch
• GitHub native – Works directly from a repo URL without manual description
• Real-time streaming – Results appear token by token, not in one blocking response
• Developer-first design – Editorial aesthetic that fits premium dev tool standards
• Deployed – Live at launchkit.kroszborg.co with Vercel edge distribution