code2prompt

code2prompt is a code ingestion tool designed to transform codebases into structured, AI-optimized prompts for LLMs. It traverses directories, builds tree structures, and gathers file information to curate context windows for tasks like code analysis, generation, and refactoring.

It is available in four primary formats:

• Core: A high-performance Rust library for code ingestion.
• CLI: A human-friendly Command Line Interface.
• SDK: A Python Software Development Kit for automation and AI agents.
• MCP: A Model Context Protocol server for providing repository access to LLMs.

code2prompt provides multiple ways to integrate context forging into your workflows, depending on your environment and needs:

• Core (Rust Library): A Rust-based library for loading, filtering, and organizing code context. It acts as a stateful context manager.
• CLI: A command-line interface for forging context from a codebase and piping it directly into prompts.
• Python API: A Python wrapper around the CLI, allowing you to use code2prompt within Python scripts and AI agents.
• MCP Server: A Model Context Protocol (MCP) server that allows LLMs to use code2prompt as a tool to forge their own context.

Code2Prompt consists of several components depending on your use case:

• Core Library: High-speed Rust library for file traversal, .gitignore respect, and Git metadata.
• CLI Tool: Human-centric tool with a minimal CLI and an interactive TUI for prompt generation and token tracking.
• Python SDK: Fast Python bindings to the Rust Core via PyPI (code2prompt-rs).
• MCP Server: A local service allowing agentic applications to read local codebases efficiently.

Code2Prompt is a modular code ingestion tool designed to help developers and AI agents interact with codebases by providing high-quality, structured context. It is available in four primary forms:

• Core: The underlying library that traverses directories, builds tree structures, and gathers file information.
• CLI: A command-line interface for humans to generate prompts, with support for Handlebars templates and clipboard/file output.
• SDK: A Python binding for the core library, ideal for AI agents and automation scripts.
• MCP: A Model Context Protocol server that runs as a local service, allowing LLMs to automatically gather codebase context.

A standard Starlight project follows this directory structure:

• src/content/docs/: Place your .md or .mdx documentation files here. Each file is automatically exposed as a route based on its filename.
• src/assets/: Store images here to be embedded in Markdown using relative links.
• public/: Place static assets like favicons here.
• src/content.config.ts: Configuration for content collections.
• astro.config.mjs: Astro configuration file.
• package.json: Project dependencies and scripts.

To open the interactive Terminal User Interface, run the code2prompt command with the --tui flag followed by the path to your codebase.

code2prompt path/to/your/codebase --tui

To create a new documentation site using the Starlight template with Astro, use the following command:

yarn create astro@latest -- --template starlight

To install the SDK for local development, clone the repository, set up a virtual environment, install development dependencies, and use maturin to build and install the package.

# Clone the repository
git clone https://github.com/mufeedvh/code2prompt.git
cd code2prompt

# Set up virtual environment and install dependencies
python3 -m venv .venv
source .venv/bin/activate
pip install maturin pytest

# Build and install locally
maturin develop -r

To ensure accurate token counting, match the encoding setting in your .c2pconfig to your target LLM's tokenizer:

• Use o200k for GPT-4o.
• Use cl100k for Claude and GPT-4.

encoding = "o200k"

After configuring your files, settings, and templates, press Enter to generate the final prompt in the Output tab.

Output Actions:

• Review: Use ↑ / ↓, PgUp, or PgDn to scroll through the generated text.
• Copy: Press c to copy the entire prompt to your system clipboard.
• Save: Press s to save the prompt to a local Markdown file.
• Regenerate: If you need to change file selections, press 1 to jump back to the Selection tab, make changes, and press Enter to regenerate.