cprp

Let's say I have this codebase that I built a few months ago, and I completely forgot what it was about. I want to send it over to Claude, ChatGPT, or DeepSeek, but it's tedious to copy and paste every file, and even if I did, the LLM would have no idea what my file structure looks like. This is what this tool aims to address.

cprp (or copyrepo) is a command-line tool designed to help users easily convert a directory into a LLM-friendly format.

This program recursively searches through a specified directory and outputs, directly to the clipboard, a directory structure, as well as the content of any non-directory files.

Installation

pip install cprp

Usage

cprp /path/to/directory

This outputs the following directly to the clipboard, as well as in the terminal:

# DIRECTORY STRUCTURE

my-project/
|-- database/
|   |-- sqlite-connect.py
|  `-- models.py
`-- main.py

## database/sqlite-connect.py
// Contents of sqlite-connect.py

## database/models.py
// Contents of models.py

## main.py
// Contents of main.py

If you want to just see the tree, you can use the --tree-only flag.

cprp --tree-only /path/to/directory

Any additional flags and commands can be viewed through the -h or --help flag.

cprp -h

Ignore behavior

Directory entries are processed alphabetically so the generated output is stable between runs. cprp applies its packaged default exclusions, such as .git, .venv, and Python cache files, followed by the .gitignore in the root of the directory being scanned. Ignore patterns are matched relative to that root directory.

Nested .gitignore files are loaded as folders are scanned. Each file's rules are relative to its own folder, and deeper rules override matching parent rules, including ! exceptions. Excluded directories are not entered, so rules inside them cannot re-include their contents.

When scanning a subfolder of a Git repository, cprp inherits parent .gitignore rules up to the nearest repository root (including worktrees with a .git file). Outside a Git repository, rule discovery starts at the selected directory. Global Git ignore settings and .git/info/exclude are not loaded.

File-content headings use paths relative to the scanned directory, such as ## database/models.py, so files with the same name in different directories remain distinguishable.

Symbolic links (or folder shortcuts) will be followed unless excluded, or specified. A link that goes back to an ancestoer directory will show up in th directory tree, but will be ignored (to prevent loops). Broken or unresolvable links are skipped.

Use cprp -ns /path/to/directory (or --no-follow-symlinks) to omit symbolic links to files and folders from both the tree and copied contents.

Requirements

• pyperclip (for copying to clipboard)
• typer (for command-line utility)
• pathspec (for gitignore parsing)

Changelog

v0.0.4

• Fixed a bug where errors were printed in the wrong format.
• Fixed a bug where invalid directories still resulted in output being printed.
• Added deterministic output and support for the scanned directory's root .gitignore.
• Changed file-content headings to use root-relative paths.

v0.0.3

• Initial release

Work in progress

• Ignore functionality
  • Custom ignores (an 'exclude.txt' that could be entered as an argument)
• Include functionality
  • Custom includes (only include a certain filetype, for example, only .py files)
• Individual files
  • Add a single file as an argument to copy its contents directly to clipboard
• Custom formatting
  • Instead of outputting to keyboard, allow outputting to file
  • Multiple format support (.json)