GitHub - context-root/code-clip

code-clip prints source files and formats them for LLM chat inputs (ChatGPT, Claude, etc.) while respecting your .gitignore, .ignore, and .cursorignore files.

Use it like this:

$ code-clip . | pbcopy
Copied 7 files (~4251 tokens)

It's designed for performance and with one thing in mind: to get the code you actually want to paste in to your LLM chat.

What it actually does:

$ code-clip .
# src/main.go

```go
package main

import "fmt"

func main() {
 fmt.Println("Hello, World!")
}
```

# src/utils/helpers.go

```go
package utils

func Add(a, b int) int {
 return a + b
}
```

Use -o xml for Claude and ChatGPT.

$ code-clip . -o xml | pbcopy
Copied 7 files (~4251 tokens)

For clipboard support, pipe the output to pbcopy (macOS), xclip (Linux), or clip (Windows).

Features

• Recursively print directory contents
• Respects .gitignore, .ignore, and .cursorignore files, even in parent directories.
• Automatically respects .gitignore, .ignore, and .cursorignore files in the current directory and all ancestor directories up to the root.
  • Ignore custom files and folders using -i or --ignore
• Output format: Markdown or XML (-o or --output-format). XML is highly recommended for Claude and ChatGPT.
• Automatically prints a token-count estimation to stderr upon completion.
• Limit the traversal depth with -d or --max-depth
• Customize the Markdown heading depth using -m or --markdown-depth (e.g., -m 3 will generate ### filename.go).

Installation

go install github.com/omarish/code-clip/cmd/code-clip@latest

(Or clone this repository and run make install) (Coming soon: homebrew and npx packages)