📦 repo-to-llm-context 📦

Stop copy-pasting code. Start shipping smarter prompts.

the ultimate context packer for your AI coding assistant. it scans your repo, ditches the junk, and bundles the good stuff into one perfect, clipboard-ready prompt.

  •  

🧭 Quick Navigation

⚡ Get Started • ✨ Key Features • 🎮 Usage & Examples • ⚙️ Advanced Flags • 🆚 Comparison

---

repo-to-llm-context is the project manager your AI assistant wishes it had. Stop feeding your LLM random files and praying for a good answer. This tool acts like a pro developer, reading your entire project, intelligently selecting the most relevant files, and packaging them into a perfectly structured prompt so your AI can actually understand what the hell is going on.

🧠 Smart Filtering Ditches node_modules & junk

🎯 Relevance Scoring Puts the important code first

📋 Clipboard Ready One command, ready to paste

How it slaps:

• You: cd my-project && context
• context: Scans, filters, scores, formats, and copies.
• You: Cmd+V into Claude/ChatGPT/Gemini.
• LLM: "Ah, I see. A well-structured project. Here is your god-tier answer."

---

💥 Why This Slaps Other Methods

Manually prepping context is a vibe-killer. repo-to-llm-context makes other methods look ancient.

❌ The Old Way (Pain)	✅ The `context` Way (Glory)
Open 15 files in VS Code. Frantically copy-paste into a text file. Realize you forgot the Dockerfile. Curse as you hit the token limit. Get a mediocre answer from a confused LLM.	cd my-project context Paste. Receive genius-level insights. Go grab a coffee. ☕

We're not just concatenating files. We're building a high-signal, low-noise prompt with intelligent depth-first traversal that processes directories systematically, prioritizes README files, and prevents massive files from breaking your LLM's context window.

---

🚀 Get Started in 60 Seconds

The context command will be available in your terminal after installation.

Platform	Recommended Method	One-liner
🍎 macOS	Homebrew	brew install yigitkonur/context/context
🪟 Windows	Scoop	scoop bucket add context https://github.com/yigitkonur/scoop-context && scoop install context
🐧 Linux	Homebrew	brew install yigitkonur/context/context

🍎 macOS: Homebrew (Recommended)

The cleanest, most native experience for Mac users.

# Add the tap and install (one-time setup)
brew tap yigitkonur/context
brew install yigitkonur/context/context

🪟 Windows: Scoop (Recommended)

Scoop is the developer's choice for Windows package management. No admin rights needed!

# First, install Scoop if you don't have it (one-time)
Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser
Invoke-RestMethod -Uri https://get.scoop.sh | Invoke-Expression

# Add the bucket and install
scoop bucket add context https://github.com/yigitkonur/scoop-context
scoop install context

🪟 Windows: pipx (Alternative)

If you already have Python installed and prefer pipx:

# 1. Install pipx if you don't have it
python -m pip install --user pipx
python -m pipx ensurepath

# 2. Install the tool
pipx install repo-to-llm-context

🐧 Linux: Homebrew (Recommended)

Homebrew works great on Linux too!

# Install Homebrew for Linux (if not installed)
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

# Add the tap and install
brew tap yigitkonur/context
brew install yigitkonur/context/context

🐧 Linux: pipx (Alternative)

For Linux users who prefer pipx over Homebrew:

# 1. Install pipx if you don't have it
python3 -m pip install --user pipx
python3 -m pipx ensurepath

# 2. Install the tool
pipx install repo-to-llm-context

✨ Zero Manual Setup: After installation, the context command should be ready to go. If not, just restart your terminal!

---

🎮 Usage: Fire and Forget

The workflow is dead simple.

1. Navigate to Your Project

cd /path/to/your/killer-app

2. Run the Command

context

Your clipboard is now loaded with perfectly formatted Markdown.

3. Paste & Prompt Go to your favorite LLM and paste the context. Now you can ask the real questions.

� GitHub Repository Support (NEW!)

You can now clone and analyze any public GitHub repository directly without downloading it locally!

Supported Formats:

• owner/repo - Shorthand format
• owner/repo@branch - Specific branch
• https://github.com/owner/repo - Full URL
• github.com/owner/repo - Short URL

Examples:

# Clone and scan a repository
context sindresorhus/is-online

# Scan specific branch
context facebook/react@main

# Full GitHub URL
context https://github.com/torvalds/linux

# With filtering options
context owner/repo --include "*.py" --preview

# Output to file instead of clipboard
context owner/repo --output analysis.md

The tool automatically:

• Performs shallow clones for speed
• Manages temporary directories (auto-cleanup on exit)
• Shows GitHub source links in output
• Applies the same intelligent filtering to remote repositories

�📏 Large File Control (NEW!)

Got massive JSON files or generated code breaking your LLM context? We've got you covered.

• Skip large files entirely:

  context --skip-large-files --max-file-chars 5000

• Truncate large files with smart preview:

  context --truncate-large-files --max-file-chars 8000

• Custom limits for different projects:

  context --max-file-chars 15000 --truncate-large-files

The tool shows exactly what it's doing:

INFO: Truncated response.json: 1,487,897 → 10,075 chars
Success: 530,745 chars copied to clipboard

Output Control 🕹️

Don't want it on your clipboard? No problem.

• Save to a file:

  context --output project_context.md

• Print directly to your terminal (for piping or peeking):

  context --stdout

---

✨ Feature Breakdown: The Secret Sauce

Feature	What It Does	Why You Care
🧠 Smart Filtering No junk allowed	Auto-excludes node_modules, venv, builds, .git, logs & more	Stops you from wasting tokens on garbage
🎯 Depth-First Sorting Perfect file order	Traverses directories systematically, README.md files first	Your LLM gets context in logical, hierarchical order
📏 Large File Control Token-aware sizing	Skip or truncate files over configurable limits (default: 50K chars)	Never blow your LLM's context window again
🏗️ Project Tree Visual context	Includes a tree-style view of what's included	The AI (and you) can see the project structure
⚙️ Git-Aware Respects your repo	Can read your .gitignore and check tracking status	Context matches your actual source code
📋 Clipboard Ready Cmd+C on steroids	Copies the entire formatted output in one go	Zero manual work between terminal and AI
🔧 Hyper-Configurable You're the boss	Flags to include/exclude anything you want	Fine-tune the context for any weird project
🔒 Privacy First No path leaks	Masks your local home directory path in the summary	Share your code, not your user folder
🎯 Interactive Mode Guided setup	Run --interactive for a step-by-step config wizard	Perfect for first-time users or complex setups

---

⚙️ Advanced Usage & Customization

The defaults are great, but you can dial it in just right.

Expand for the full list of command-line flags

Filtering and Inclusion Control

• --include PATTERN: Glob pattern to force inclusion of files/directories that might be excluded (e.g., --include "config/**.yaml").
• --exclude PATTERN: Glob pattern to add custom exclusions beyond the defaults (e.g., --exclude "*.log").
• --include-only: A powerful mode that includes only files matching --include patterns, excluding everything else.
• --exclude-extension EXT: Exclude all files with a specific extension (e.g., --exclude-extension .tmp).
• --include-extension EXT: Force include files with an extension that is normally excluded by default.

Override Default File Type Exclusions

• --include-json: Include .json / .jsonc files.
• --include-yaml: Include .yaml / .yml files.
• --include-xml: Include .xml files.
• --include-html: Include .html / .htm files.
• --include-css: Include .css files.
• --include-sql: Include .sql files.
• --include-csv: Include .csv / .tsv files.
• --include-markdown: Include all Markdown files, not just the root README.md.

Size and Content Control

• --max-size SIZE: Exclude files larger than the specified size (e.g., 500k, 10M). Default is 2M.
• --max-file-chars N: Set maximum characters per file (default: 50,000). Works with skip/truncate options.
• --skip-large-files: Skip files that exceed the --max-file-chars limit entirely.
• --truncate-large-files: Keep large files but show only the first N characters with a truncation notice.
• --include-binary: Attempt to include files detected as binary (default is to exclude them).
• --max-depth N: Limit scanning to a maximum directory depth.

Git Integration Behavior

• --no-gitignore: Ignore .gitignore rules and Git tracking status entirely.
• --gitignore-only: (Default) Use .gitignore rules for exclusion but don't filter based on Git tracking status.
• --use-git: Use both .gitignore rules and only include files that are tracked by Git.

Output and Execution Behavior

• --output FILE: Write output to a file instead of the clipboard.
• --stdout: Print the full output to the terminal.
• --no-clipboard: Disable automatic copying to the clipboard (useful when using --stdout or --output).
• --preview: Show a summary of what would be included without processing files or generating output.
• --dry-run: Run the entire process but do not write any output to the clipboard, file, or stdout.
• --sort-alpha: Override the relevance-based sorting and sort files alphabetically instead.

Information

• --version: Display the current version and exit.
• --check-updates: Check for available updates.

---

🆚 Why This Slaps Other Methods

Method	The Pain 😩	The context Way 😎
Manual Copy/Paste	You'll miss a file. You'll include junk. You'll hate your life.	One command. Perfect context. Every time.
cat file1 file2 > out.txt	Zero structure. No filtering. Still manual. Basically useless.	Auto-filters, adds a file tree, and formats beautifully.
Sharing a GitHub Link	LLM can't see local changes. Can't access private repos.	Works offline. Works on your latest, unpushed code.
Simple tree command	Shows structure but includes zero code content.	Gives you the full package: structure AND content.

---

🛠️ For Developers & Tinkerers

Running from Source

Want to hack on the code? Easy.

1. Clone the repo:

   git clone https://github.com/yigitkonur/code-to-clipboard-for-llms.git
   cd code-to-clipboard-for-llms

2. Set up a virtual environment and install in editable mode:

   python3 -m venv venv
   source venv/bin/activate
   pip install -e .

   Now, any change you make to llmcontext.py will be live on your context command.

Fork & Customize

If you fork the repo, you can permanently change the default filters by editing the constants at the top of llmcontext.py.

---

🔥 Common Issues & Quick Fixes

Expand for troubleshooting tips

Problem	Solution
context: command not found	Restart your terminal. 99% of the time, this is the fix. If not, run pipx ensurepath (for pipx) or check your PATH environment variable.
Clipboard isn't working	Linux users: You might need a clipboard utility. Run sudo apt install xclip or sudo pacman -S xclip. For any OS, you can always use --stdout or --output my_context.md to bypass the clipboard.
.gitignore is ignored	Make sure you have gitignore-parser installed. The tool uses .gitignore by default. Use --no-gitignore to disable.
Script errors out	Make sure you're on Python 3.8 or newer (python3 --version).

Windows-specific issues:

Problem	Solution
Scoop install fails	Make sure you've enabled script execution: Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser
context not recognized in PowerShell	Close and reopen PowerShell, or run scoop reset context to refresh the shim.
Permission denied errors	Scoop installs to user directory, so admin rights shouldn't be needed. If using pipx, run PowerShell as Administrator.
Python not found (pipx method)	Download Python from python.org and ensure "Add Python to PATH" is checked during installation.

---

Built with 🔥 because manually crafting LLM prompts is a soul-crushing waste of time.