ctx: AI-Ready Codebase Context

AI coding assistants lack architectural context. Every time I asked Claude or ChatGPT to write code, I’d waste time explaining my codebase structure, patterns, and conventions.

I needed a lightweight, always-fresh representation of my codebase—just the skeleton, no implementation details. So I built ctx.

What It Does

ctx extracts your codebase into .ctx/ with “skeletons”—function signatures, class structures, type definitions, exports. Roughly 6% of the original code size. Feed these to your AI, and it understands your entire architecture before touching real files.

The Workflow

go install github.com/dakshpareek/ctx@latest
cd /path/to/your/project
ctx init          # bootstrap .ctx/ and seed the index
ctx ask           # auto-sync + prompt to .ctx/prompt.md
# ... share prompt, save skeletons to .ctx/skeletons/ ...
ctx update        # mark skeletons current after AI saves them
ctx bundle        # optional: export .ctx/context.md for pairing

1. ctx ask scans for changes and writes an AI prompt to .ctx/prompt.md
2. Paste prompt into Claude/ChatGPT
3. AI generates skeletons → save to .ctx/skeletons/
4. Run ctx update
5. Next time, only new/changed files need skeletons

Why Manual First

Phase 1 is intentionally manual:

• Simple and reliable
• Works with any LLM
• No vendor lock-in
• Build incrementally

Phase 2 adds automation, watch mode, git hooks.

Get Started

go install github.com/dakshpareek/ctx@latest
cd /path/to/your/project
ctx init
ctx ask

GitHub | Docs