packmd

Instant AI context. PackMD is a command‑line tool that converts GitHub repos, local directories, and live web pages into clean, token-efficient Markdown digests optimized for ChatGPT, Claude, and other LLMs.

Installation

Global (recommended)

npm install -g packmd

Now you can use packmd from anywhere:

packmd --help

Run without installing (using npx)

npx packmd <target> [options]

Quick Start

# Generate a digest from a public GitHub repo
packmd [https://github.com/vercel/next.js](https://github.com/vercel/next.js) -o next.md

# Scrape a documentation website (requires Jina API key for higher limits)
packmd [https://react.dev](https://react.dev) --jina-api-key YOUR_KEY --copy

# Digest the current directory (respects .gitignore)
packmd .

# Digest a local folder with custom options
packmd ~/projects/my-app -m 300 -s 50 --exclude "*.test.js"

Usage

packmd [target] [options]

| Argument | Description | | -------- | --------------------------------------------------------------------------------------------------------------------------------------------------- | | target | GitHub URL (e.g., https://github.com/owner/repo), owner/repo slug, local directory path, or any webpage URL. Defaults to current directory (.). |

Options

| Flag | Description | | ----------------------------- | ------------------------------------------------------------------------------------------ | | -o, --output <path> | Save the generated Markdown to a specific file (e.g., digest.md). | | -c, --copy | Copy the output directly to your clipboard (no file saved). | | -t, --token <token> | GitHub Personal Access Token for private repositories or higher rate limits. | | -m, --max-files <number> | Maximum number of files to include (default: 200). | | -s, --max-file-size <kb> | Maximum file size in KB (files larger are skipped) (default: 100). | | -i, --include <patterns...> | Glob patterns to explicitly include (e.g., "*.ts" "*.md"). | | -e, --exclude <patterns...> | Glob patterns to explicitly ignore (e.g., "test/**" "*.log"). | | --jina-api-key <key> | Jina Reader API key – raises the free rate limit from 20 to 500 requests per minute. | | --no-gitignore | Ignore .gitignore rules when scanning a local directory (by default they are respected). | | -v, --version | Show the version number. | | -h, --help | Show help. |

Note: When using --output, if the file already exists, you will be prompted to overwrite or enter a new name.

Examples

GitHub repository

# Basic
packmd facebook/react -o react.md

# With token (for private repos)
packmd my-org/private-repo -t ghp_xxxxx -o private.md

# Limit to 300 files, exclude everything in the "examples" folder
packmd vercel/next.js -m 300 -e "examples/**" -o next-limited.md

Web page (via Jina Reader)

# Simple scrape
packmd [https://docs.nestjs.com](https://docs.nestjs.com) --copy

# With Jina API key (higher rate limit)
packmd [https://tailwindcss.com/docs](https://tailwindcss.com/docs) --jina-api-key jina_xxxxx -o tailwind.md

Local directory

# Digest the current folder
packmd .

# Digest a specific path
packmd ~/code/my-project -o project.md

# Override .gitignore and include only TypeScript files
packmd ./src --no-gitignore -i "*.ts" "*.tsx"

Combine options

packmd [https://github.com/expressjs/express](https://github.com/expressjs/express) -o express.md -m 100 -s 50 -e "test/**" "*.spec.js"

How It Works

The CLI uses the same core engine (@packmd/core) as the web app:

1. Input detection – determines if the target is a GitHub repo, a web page, or a local path.
2. Fetching – for GitHub, uses the GitHub API; for web pages, uses Jina Reader; for local, uses Node.js fs.
3. Filtering – applies ignore rules, globs, size limits, and binary detection.
4. Processing – builds a directory tree and downloads file contents (in parallel for GitHub).
5. Markdown generation – creates a header with metadata, an ASCII tree, and fenced code blocks for each file.

Progress is shown via a live spinner (ora). Output can be saved to a file or copied to your clipboard.

Notes

• GitHub rate limits – Without a token, you are limited to 60 requests per hour. Use -t to raise this to 5,000 per hour.
• Jina rate limits – Without an API key, you get 20 requests per minute. Provide --jina-api-key for 500 RPM.
• The CLI respects .gitignore by default. Use --no-gitignore to scan everything.
• Binary files (images, fonts, archives) are automatically skipped.

License

MIT © Holiday