I built a CLI that gives any AI instant context about your project

Source: Dev.to

Overview

Every time I start a new AI session I spend the first few minutes explaining the same things:

• “This is a FastAPI project”
• “We use SQLAlchemy for the ORM”
• “The main entry point is src/api/main.py”
• “Recent work has been on the auth module”

It’s tedious, and AI tools like Claude Code, ChatGPT, and Gemini start cold each session.
To solve this, I created ctx, a small CLI that builds a Markdown context pack for any project and injects it into an AI chat instantly.

Installation

pip install ctx

Creating a Context Pack

ctx save myproject          # scan project, save as a context pack

ctx save automatically:

• Stack detection – finds pyproject.toml, package.json, Cargo.toml, go.mod, Gemfile, etc.
• Structure map – directory tree of src/, tests/, api/ folders
• Git log – last 10 commits so the AI knows recent work
• README summary – first few lines as project context
• Your notes – optional extra information

The result is a clean Markdown file that any AI can parse immediately.

Example Pack (myproject.md)

# myproject

## Stack
- Python
- Detected from: pyproject.toml

## Structure
src/
  api/
  models/
tests/

## Recent commits
- feat: add user auth
- fix: resolve migration conflict
- refactor: extract service layer

## README
MyProject is a FastAPI app for managing...

## Notes
Main entry: src/api/main.py
Auth lives in src/auth/ — JWT-based, no sessions

Injecting the Context

ctx inject myproject                      # → clipboard (paste into ChatGPT, Gemini, etc.)
ctx inject myproject --target claude      # → writes CLAUDE.md in current directory
ctx inject myproject --target chatgpt     # → clipboard, formatted as a system prompt

• Claude Code reads CLAUDE.md automatically when you open a project—no manual paste needed.
• For ChatGPT, Gemini, or any other model, a single paste at the start of the session fully loads the context.

Scope Options

ctx save myproject --scope global   # → ~/.ctx/packs/myproject.md (default, usable from any directory)
ctx save myproject --scope local    # → .ctx/myproject.md (project‑specific, git‑committable)

Local packs take priority. Commit the .ctx/ directory to your repository and the whole team gets the same context.

CLI Reference

Why “Context Engineering”?

“Context engineering” is the new prompt engineering.
Models are powerful, but they’re limited when they lack sufficient context about your project—your conventions, current work, and architectural decisions.

ctx provides a local, zero‑dependency way to supply that context:

• No account required
• No sync service
• Pure Markdown files you control

Usage in Any Project

ctx save myproject --notes "Add anything you want the AI to know"
ctx inject myproject --target claude

Source

github.com/LakshmiSravyaVedantham/ctx