ClawDocs

Generate AI-agent-optimized documentation for any codebase using Claude Code.

ClawDocs creates a comprehensive prompt that Claude Code executes to analyze your project and generate documentation following the Diataxis framework combined with Anthropic's agent-development best practices.

Prerequisites

• Claude Code installed and configured

Installation

npm install -g clawdocs

Or use directly with npx:

npx clawdocs generate

Usage

Generate Documentation Prompt

# Navigate to your project
cd your-project

# Generate the Claude Code prompt
clawdocs generate

# Run the generated prompt with Claude Code
claude ".clawdocs/generate-docs.md"

Options

clawdocs generate [options]

Options:
  -d, --depth <level>   Documentation depth: minimal, standard, comprehensive (default: "standard")
  -o, --output <path>   Output path for prompt file (default: ".clawdocs/generate-docs.md")
  --run                 Immediately run the prompt with Claude Code after generation

Depth Levels

minimal - Essential files only:

• CLAUDE.md
• docs/README.md
• docs/03-reference/agent-reference/codebase-map.md

standard (default) - Balanced documentation:

• All configuration files (CLAUDE.md, AGENTS.md, CLAUDE.LOCAL.md)
• Core docs/ structure
• Codebase map and architecture
• Basic how-to guides
• Agent workflows

comprehensive - Complete documentation:

• Everything in standard
• Detailed tutorials
• Full API reference
• Architecture Decision Records (ADRs)
• Troubleshooting guides
• Security and performance docs

What Gets Generated

When you run the prompt with Claude Code, it analyzes your codebase and generates:

Root Files

• CLAUDE.md - Project configuration for Claude Code (commit this)
• CLAUDE.LOCAL.md - Personal working memory (git-ignored)
• AGENTS.md - Agent skills framework with capabilities

docs/ Structure

docs/
├── README.md                    # Master index
├── META-AGENT-PROTOCOLS.md      # Documentation maintenance rules
├── 00-getting-started/          # Installation and setup
├── 01-tutorials/                # Step-by-step learning
├── 02-how-to-guides/            # Task-oriented recipes
├── 03-reference/                # Technical specifications
│   └── agent-reference/
│       └── codebase-map.md      # Complete codebase map
├── 04-explanation/              # Architecture and concepts
├── 05-agent-workflows/          # Executable YAML workflows
├── 06-agent-guardrails/         # Restrictions and standards
├── 07-development/              # Dev environment setup
├── 08-deployment/               # Deployment procedures
├── 09-adrs/                     # Architecture decisions
├── 10-conventions/              # Coding standards
└── 11-troubleshooting/          # Problem-solving guides

Supported Tech Stacks

ClawDocs auto-detects your stack to provide context to Claude:

Backend:

• Go (Gin, Echo, Fiber)
• Node.js (Express, Fastify, NestJS, Hono)
• Python (FastAPI, Django, Flask)
• Rust (Actix, Axum)

Frontend:

• React, React Native (Expo)
• Next.js, Vue, Svelte

Infrastructure:

• Docker, Docker Compose
• PostgreSQL, MySQL, MongoDB

Monorepos:

• npm/yarn workspaces
• Lerna, pnpm workspaces

Documentation Philosophy

ClawDocs follows these principles:

1. Token Efficiency - Informative but concise. No emojis or decorative elements.
2. Agent-Optimized - Structured for AI agent consumption with clear navigation.
3. Living Documentation - Meta-protocols for keeping docs in sync with code.
4. Just-in-Time Context - Progressive disclosure: indexes first, details on demand.
5. Diataxis Framework - Separates tutorials, how-tos, reference, and explanation.

Example

# Generate and immediately run
clawdocs generate --depth comprehensive --run

# Or step by step
clawdocs generate --depth standard
claude ".clawdocs/generate-docs.md"

Contributing

Contributions welcome. Please open an issue or PR on GitHub.

License

MIT