dot.ai

AI-powered code generation from plain Markdown specifications

What is dot.ai?

Write your requirements in .ai files (plain Markdown), run dot gen, and let AI coding agents implement them. Perfect for rapid prototyping, maintaining consistency, and automating repetitive coding tasks.

Key Features

• Plain Markdown - No frontmatter, just write specifications
• CLI-First - All configuration via command-line flags
• Multiple Agents - Support for claude-code, cursor, aider
• Smart Updates - Hash-based change detection for incremental generation
• Parallel Processing - ~5x faster for multiple files
• Type Safe - Built with TypeScript, comprehensive test coverage

Quick Start

Installation

npm install -g @dean0x/dot

Authentication

dot.ai uses Claude Code under the hood. You can authenticate two ways:

Option 1: Claude Max/Pro Plan (Recommended)

• Claude Code automatically uses your browser authentication
• No API key needed

Option 2: API Key

export ANTHROPIC_API_KEY=your_api_key_here

Get your key from: https://console.anthropic.com/

Usage

1. Create a .ai file (plain Markdown):

# Button Component

Create a reusable React Button component with TypeScript.

## Requirements
- Support variants: primary, secondary, danger
- Include loading and disabled states
- Full test coverage

2. Generate code:

dot gen

That's it! The .dotai/ directory is created automatically, artifacts are generated, and state is tracked.

3. Make changes - Edit your .ai file and run dot gen again. Only changed files are processed.

CLI Commands

dot gen [path]           # Generate code from .ai files
dot status [path]        # Show which files have changed
dot ls [path]            # List all .ai files and artifacts
dot clean                # Clear all generation state

Configuration

All configuration via command-line flags:

# Select agent
dot gen --agent claude-code

# Enable iterative mode (agent can update spec and re-run)
dot gen --iterate

# Set iteration limit when using --iterate
dot gen --iterate --max-iterations 5

# Enable parallel processing
dot gen --parallel --concurrency 10

# Force regenerate all files
dot gen --force

Run dot gen --help for all options.

How it Works

1. Write Specifications - Create .ai files with your requirements in Markdown
2. Smart Detection - dot.ai uses hash-based change detection to find what changed
3. Agent Invocation - Your chosen coding agent implements the specification
4. Artifact Tracking - Generated files are tracked for future incremental updates
5. Iterate - Edit specs, run dot gen, repeat

Iterative Processing

dot.ai supports iterative mode with the --iterate flag - after generating artifacts, it re-reads the .ai file to see if the agent updated it with new tasks. This enables self-directed refinement workflows where the agent can add new tasks for itself.

Enable with --iterate and control iterations with --max-iterations.

Examples

See the examples/ directory for:

• Self-reflecting agents
• Recursive processing patterns
• Multi-file project generation

Documentation

• Architecture - System design and patterns
• Contributing - Development guide
• Specification - Full technical specification
• Changelog - Version history

Requirements

• Node.js >=20.0.0
• Claude Code CLI (install separately: https://github.com/anthropics/claude-code)

Security

• Agent whitelist prevents arbitrary code execution
• Path traversal protection
• Input validation and range enforcement
• Dangerous flags blacklisted

See Security Policy for details.

License

MIT © dean0x

Support

• Issues - Bug reports and feature requests
• Discussions - Questions and community

Built with Claude Code 🤖