// Documentation

How PackMD Works

PackMD converts codebases, local directories, and live web pages into clean, token‑efficient Markdown for LLMs.

Overview & Engine

Behind the scenes, PackMD combines high‑performance filesystem traversal with modern web scraping APIs. It strips noise, parses syntax, and formats everything into a unified context window.

Repositories & Local

Scans projects, respects .gitignore, and generates structured Markdown.

Web Pages

Converts web pages into clean Markdown by removing unnecessary page content.

CLI Usage

Run PackMD instantly from your terminal or plug it into our web interface for live previews and instant clipboard copying.

Installation
bun add -g packmd
Usage
~terminal
# Pack a GitHub repository or remote web page into clean Markdown packmd https://github.com/thelastofinusa/packmd -o digest-pack.md -m 300 packmd https://packmd.vercel.app --copy
Local Directories & Advanced Usage

PackMD isn't limited to remote URLs — point it at any local directory, and combine flags to fine-tune exactly what gets included.

~terminal
# Digest the current directory packmd . # Combine advanced options: cap files, exclude tests, copy to clipboard packmd facebook/react --max-files 300 --exclude "*.test.js" --copy # Skip .gitignore rules when scanning a local directory packmd . --no-gitignore
CLI Options & Flags
DescriptionFlagAlias
output

Specify output file path (e.g., digest.md)

--output-o
copy

Copy output directly to clipboard upon completion

--copy-c
token

GitHub Personal Access Token for private repos / higher rate limits

--token-t
max-files

Limit total number of files to include (default: 200)

--max-files-m
max-file-size

Maximum allowed size (in KB) for individual files (default: 100)

--max-file-size-s
include

Glob patterns to explicitly include files or directories

--include-i
exclude

Glob patterns to explicitly ignore files or directories

--exclude-e
no-gitignore

Don't respect .gitignore rules when scanning a local directory

--no-gitignore

Processing Pipeline

Identifies whether the input is a GitHub repository or a standard webpage.

FAQ & Troubleshooting

Unauthenticated GitHub requests are capped at 60/hour. Pass a Personal Access Token with --token to raise this to 5,000/hour.