An AI that reads your code
Creating docs, likeyour code.
Drop a GitHub URL or describe your project. Caveman scans your file tree, maps dependencies, and writes a README that matches what you actually built.
Trusted by teams building on the world's top platforms
What we do
We don't template. We read your code.
UI/UX design isn't just about aesthetics. Neither is documentation - it's about making interactions smooth and workflows clear. Caveman turns your codebase into a polished, accurate document in seconds.
Deep scanning
Reads your actual source
Caveman walks your file tree, resolves imports, and detects your framework, language, and architecture - no guessing, no generic filler.
Adaptive sections
Sections that match your stack
Installation, usage, API reference, environment variables - only included when your project actually has them.
README sections
writing tones
documentation depths
target generation window
How it works
Three cave-simple steps
From URL to polished README in under 90 seconds.
Sign in & start
Continue with Google to open the README generator.
Your account tracks the early-access generation limit. No credit card is needed.
Connect your project
Paste a GitHub URL or type a quick description. Zero setup.
Public repositories work by default. Private scans require a future user-scoped GitHub authorization flow.
Generate & refine
Caveman scans the project and writes a grounded README.
Preview, edit, copy, or download the generated Markdown before you ship it.
Features
Every feature, no filler
Template generators fill in your project name. Caveman reads your entire codebase.
Deep code analysis
Reads your actual source files - entry points, components, APIs, and config. Writes docs that match your real codebase.
Stack auto-detection
Detects React, Next.js, Tailwind, Prisma, tRPC, Express, Django, and more from your dependencies and file tree.
Package manager aware
Reads lockfiles to generate accurate install commands - npm, pnpm, yarn, bun. Even handles monorepo workspaces.
Architecture mapping
Identifies your project structure, component hierarchy, API routes, and data models. Generates accurate architecture diagrams.
Health scoring
Built-in README quality checker scores your docs against 11 criteria. Missing sections? It auto-generates them.
17+ customizable sections
Installation, Usage, API Docs, Contributing, Features, Architecture, Testing, Deployment - toggle any combination.
Tone & style control
Switch between Technical, Friendly, or Enterprise tone. Choose Minimal, Standard, or Comprehensive depth. One click.
Under 90 seconds
Designed to turn a repository or project description into a polished README in under 90 seconds, with preview, editing, copy, and download.
GitHub deep scan
Traverses up to 3 directory levels and selects a bounded set of high-signal source files - components, API routes, config, and schemas - before writing.
Recent work
READMEs that actually look good
No template filler. Every section is written from your actual source code.
## Installation
```bash
npm install next-forge
```
## Usage
```tsx
import { ForgeProvider, useForge } from "next-forge";
function App() {
const { data, isLoading } = useForge("/api/build");
return <div>{isLoading ? "Building..." : data}</div>;
}
```## Tech Stack - **Runtime:** Node.js 20+, Bun 1.0+ - **Framework:** Next.js 15 with App Router - **Database:** PostgreSQL 16 via Prisma ORM - **Auth:** NextAuth v5 with Google provider
## API Reference ### `process_pipeline(data: dict) -> dict` Processes raw data through the transformation pipeline. | Parameter | Type | Description | |-----------|------|-------------| | data | dict | Input data with `source` and `config` keys | | normalize | bool | Whether to normalize output (default: True) |
What you get
Production-ready. Every time.
LaunchKit
A typed starter for shipping production-ready web apps with auth, billing, and background jobs.
LaunchKit gives teams a typed foundation for building and shipping web products without repeating the same setup work.
Comparison
Why template generators fall short
Testimonials
Our clients say it best
"I spent more time writing READMEs than code. Caveman dropped that to zero. It actually reads my components and writes docs that match what I built."
Alex Chen
Founder, Finch
"The deep scan is what sold me. It found my tRPC router, my Prisma schema, and my Tailwind config - the README it wrote was better than what I would have written."
Jordan Taylor
Engineer, Rivet
"We were using a template generator before. This is different - it understands the architecture, not just the name of the project. Night and day."
Samir Patel
Lead Dev, Mantle
FAQ
Questions? We got answers.
Yes. It fetches your file tree from GitHub, identifies entry points, and sends a bounded set of high-signal source snippets to the AI. It reads your components, APIs, config files, and dependencies to understand what you built.
Let's talk.
One click. Your repo scanned. A README that describes what you actually built - not what a template thinks you built.