Code Context Builder

A powerful, keyboard-driven desktop application for creating high-quality, LLM-optimized prompts from your codebase. Designed for maximum efficiency with intuitive hotkeys and a streamlined workflow. Built with Tauri, React, and Rust.

Download

Pre-built Binaries (Recommended)

Download the latest release for your platform:

• Windows: Download .exe from latest release
• Linux: Download .deb or .AppImage from latest release
• macOS: Download .dmg from latest release

All platforms are built automatically via GitHub Actions.

Build from Source

See Installation section below for build instructions.

Why I Built This

Because manually copying code into LLMs is tedious, and with expensive pro models, every token counts.

This tool helps you curate precise context, track token usage, and work at keyboard speed when you need to feed your code to any chat interface.

Features

• Smart Project Scanning: Efficiently scans project folders with customizable ignore patterns
• Interactive File Tree: Browse, search, and select files with detailed stats (lines, tokens, size)
• Multiple Output Formats:
  • Sentinel - Loud, unambiguous markers (recommended for LLMs)
  • Markdown - YAML front matter with unique fences
  • XML - Well-formed XML documents (recommended for tooling)
  • Raw - Unformatted concatenation
• Smart Code Compression: Optional compression using tree-sitter for Python and TypeScript/TSX
  • Removes comments and collapses function bodies while preserving structure
  • Reduces token count for large codebases
• Prompt Presets: Save and reuse common prompt structures with customizable pre/post prompts
• Real-time File Monitoring: Detects file changes and prompts for re-scanning
• Token Counting: Accurate token count estimation using tiktoken
• Customizable Format Instructions: Edit default instruction templates for each output format
• Project Management: Manage multiple projects with per-project ignore patterns
• Dark/Light Themes: System-aware theme support
• Keyboard-Driven Workflow: Comprehensive hotkey support for lightning-fast operation without touching the mouse

Keyboard-Driven Workflow

Code Context Builder is designed for efficiency with comprehensive keyboard shortcuts that let you work at the speed of thought:

Essential Hotkeys

Shortcut	Action
Ctrl+F	Focus search (quickly find files)
Ctrl+A	Select all files in the tree
Ctrl+Shift+A	Deselect all files
Ctrl+Shift+C	Copy final prompt to clipboard (most-used action!)
Ctrl+Shift+R	Re-scan current project
Ctrl+Shift+M	Cycle output format (Markdown → Sentinel → XML → Raw)
Ctrl+Shift+T	Toggle "Prepend Tree" option
Ctrl+↓	Expand tree level (Ctrl+Click for all levels)
Ctrl+↑	Collapse tree level (Ctrl+Click for all levels)

Search Navigation

When searching for files:

• ↓ / ↑ - Navigate through search results
• Enter - Toggle selection of highlighted file
• Esc - Clear search and unfocus

Intuitive Workflow Example

1. Ctrl+F → Type filename → Enter to select
2. Repeat for all needed files
3. Edit prompts in the sidebar
4. Ctrl+Shift+C to copy the complete, formatted prompt

No mouse required! The entire workflow can be completed with just the keyboard, making it incredibly fast once you learn the shortcuts.

Screenshots

📁 File Tree & Output Formats

Select files from your project tree and see the formatted output in real-time. The interface shows file stats (lines, tokens, size) and supports multiple output formats.

🎯 Prompt Presets

Save and reuse common prompt structures. Create custom pre-prompts (instructions) and post-prompts (tasks) for different LLM workflows.

🔄 Real-Time File Monitoring

The app detects when files have been modified since the last scan and alerts you with a visual indicator, ensuring your context is always up-to-date.

⚙️ Customizable Ignore Patterns

Use .gitignore syntax to exclude files and directories from your scans. Perfect for filtering out dependencies, build artifacts, and other non-essential files.

Installation

Prerequisites

• Node.js (v18 or higher)
• Rust (latest stable)
• npm or yarn

Build from Source

1. Clone the repository:

git clone https://github.com/jatoran/code_context_builder.git
cd code_context_builder

2. Install dependencies:

npm install

3. Run in development mode:

npm run tauri dev

4. Build for production:

npm run tauri build

The built application will be in src-tauri/target/release/.

Usage

Creating a Project

1. Click "New Project" in the Project Manager
2. Enter a project title and select the root folder
3. (Optional) Add custom ignore patterns using .gitignore syntax
4. Click "Scan" to index the project

Building Prompts (The Fast Way)

Quick workflow:

1. Ctrl+F to search for files → Enter to select
2. Expand the "Prompts & Presets" section
3. Ctrl+Shift+M to cycle through formats if needed
4. Write your Pre-Prompt (instructions) and Post-Prompt (task), or load a preset
5. Ctrl+Shift+C to copy the complete formatted prompt

Traditional workflow:

1. Browse and click files in the tree (or use search with Ctrl+F)
2. Expand the "Prompts & Presets" section
3. Choose a format (Sentinel, Markdown, XML, or Raw)
4. Either:
   • Load a saved preset, or
   • Write custom Pre-Prompt (instructions) and Post-Prompt (task)
5. Toggle options:
   • Prepend Tree: Include file tree visualization
   • Include Format Instructions: Add format explanation
   • Smart Compression: Enable code compression (Python/TS/TSX only)
6. Copy the final prompt from the preview (or press Ctrl+Shift+C)

💡 Pro Tip: Once you learn the hotkeys, you can go from opening the app to having a complete prompt on your clipboard in under 10 seconds!

Architecture

Code Context Builder uses a modern tech stack:

• Frontend: React + TypeScript + Vite
• Backend: Rust + Tauri
• Database: SQLite (stored next to executable)
• Parsing: tree-sitter for code compression
• Tokenization: tiktoken-rs for accurate token counting

See DOCUMENTATION.md for detailed technical documentation.

Contributing

Contributions are welcome! Please read CONTRIBUTING.md for guidelines.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Built with Tauri
• Code parsing with tree-sitter
• Token counting with tiktoken-rs

Support

If you encounter any issues or have questions, please open an issue.