ContextCraft

A Python CLI tool that analyzes a code repository and generates a structured Context Pack document optimized for injecting into AI coding tools (Claude, GPT, Cursor, Copilot) so they understand the codebase before writing code.

Installation

CLI

pip install llm-codepac

VS Code Extension

Coming soon on the VS Code Marketplace.

Usage

export ANTHROPIC_API_KEY=your_key_here
contextcraft init /path/to/your/repo

Environment

Create a .env file (see .env.example) or set:

• ANTHROPIC_API_KEY — required for AI synthesis (omit or use --no-ai for offline)

Output

• Default: context.pack.md in the repo root with a Quick Copy section and full Context Pack.
• JSON: Raw analysis or pack as JSON when --format json is used.

Requirements

• Python 3.11+
• Git (optional; used for recent commits and hotspot analysis)

Distribution

• Install from PyPI: pip install llm-codepac
• Install the VS Code extension: ContextCraft on the Marketplace (when published), or install manually: download the .vsix from Releases and run Extensions: Install from VSIX... in VS Code.
• Publish a new Python release: tag with git tag v0.x.x && git push --tags (triggers PyPI publish).
• Publish a new extension release: tag with git tag ext-v0.x.x && git push --tags (triggers VS Code Marketplace publish and uploads .vsix to the GitHub Release).