neuro-forge 🧠🔨

The Ultimate AI Context Compressor. Mind-meld your codebase with LLMs instantly.

When you hit a bug spanning 5 different files, you manually cat or copy-paste each one into Claude or ChatGPT. That's tedious, slow, and wastes tokens on boilerplate. neuro-forge fixes this — one command compresses your active Git context into a single optimized Markdown block and injects it straight into your clipboard.

Features

• 🧠 Analyzes your active Git tree to find exactly what you're working on
• 📊 Real token counting — shows exact token count using GPT-4 tokenizer
• ⚠️ Context limit warnings — alerts when you exceed GPT-4 or Claude limits
• ✂️ Strips dead weight (empty lines, massive breaks) to save tokens
• 📋 Injects the result directly into your OS clipboard
• 🎯 Target specific files/folders — bypass Git and pack exactly what you want
• 🕐 Time-based selection — grab files changed since "2 days ago" or "last Monday"
• 💬 Append custom prompts — include your question in the same payload
• 📁 Output to file — write to a file instead of clipboard
• ⚡ Skips node_modules, .gitignore paths, locks, and binary files
• 🖥️ Works on macOS, Windows, and Linux

Quick Start

npx @dinakars777/neuro-forge compress

Or install globally:

npm install -g @dinakars777/neuro-forge
cd my-project
forge compress

Advanced Usage

Target specific files or folders

forge compress --files src/components/Auth src/api/users.ts

Get files changed in the last few days

forge compress --since "2 days ago"
forge compress --since "last Monday"

Include a custom prompt

forge compress --prompt "Find the bug causing the null pointer exception"

The prompt will be appended to the context, so you can paste everything in one shot.

Write to a file instead of clipboard

forge compress --out context.md

Combine options

forge compress --since "3 days ago" --prompt "Review this feature" --out review.md

How It Works

1. Runs git diff --name-only and git ls-files to find your active context (or uses --since/--files if specified)
2. Strips noise and bundles file contents into a clean Markdown block:

   ## File: src/components/Button.tsx
   // ...content...

3. Calculates exact token count using GPT-4 tokenizer
4. Warns you if you're exceeding context limits (32K for GPT-4 Turbo, 128K for Claude)
5. Copies the entire payload to your clipboard — just Cmd+V into your LLM

Options

| Flag | Description | Example | |---|---|---| | --since <time> | Get files changed since a specific time | --since "2 days ago" | | --files <paths...> | Target specific files or folders | --files src/auth src/db/schema.ts | | --prompt <text> | Append a custom prompt to the context | --prompt "Explain this code" | | --out <file> | Write output to a file instead of clipboard | --out context.md |

Tech Stack

| Package | Purpose | |---|---| | execa | Fast Git command execution | | clipboardy | Cross-platform clipboard access | | js-tiktoken | Accurate GPT-4 token counting | | @clack/prompts | Beautiful CLI UI | | TypeScript | Type-safe implementation |

Why neuro-forge?

Unlike generic repo packers, neuro-forge is laser-focused on the active development workflow:

• Zero config — just run forge compress and it figures out what's relevant
• Token awareness — shows exact counts and warns about context limits
• Workflow integration — clipboard-first, prompt appending, time-based selection
• Smart defaults — targets what you're working on right now, not the entire repo

License

MIT