Jay
4.9 KiB
Connect AI
100% Local · 100% Offline · 100% Free
Your AI coding agent that lives entirely on your machine.
Overview
Connect AI is an agentic AI coding assistant that runs entirely on your local machine — no cloud, no API keys, no data leaves your computer. It reads your project, creates files, edits code, manages directories, and executes terminal commands — all through natural conversation.
Built for VS Code, Cursor, and Antigravity.
⚡ Agent Capabilities
Connect AI doesn't just answer questions — it acts. Seven built-in agent actions give it full control over your development environment:
| Action | Description |
|---|---|
| 📄 Create Files | Generates new files and directories in your workspace |
| ✏️ Edit Files | Finds and replaces specific code in existing files |
| 🗑️ Delete Files | Removes files and folders |
| 📖 Read Files | Reads any file in your workspace to understand context |
| 📂 Browse Directories | Lists contents of any subdirectory |
| 🖥️ Run Terminal Commands | Executes CLI commands (install, build, run, deploy) |
| 🧠 Second Brain | Queries your personal knowledge base (GitHub-synced) |
How It Works
You: "React로 카운터 앱 만들어줘"
Connect AI:
✅ 생성: src/App.jsx
✅ 생성: src/index.js
✅ 생성: index.html
🖥️ 실행: npm install react react-dom
All files are created directly in your local workspace — no copy-paste needed.
🎨 Interface Features
- 🔄 Real-time Streaming — Token-by-token response rendering
- ⬛ Abort (Stop) — Interrupt generation mid-stream
- 🔄 재생성 — Regenerate any response with one click
- 📎 Multimodal Input — Paste images (Cmd+V) or attach files (+)
- 💡 Syntax Highlighting — Cinematic code block rendering
- ⏱️ Thinking Bar — Visual latency indicator during inference
📥 Installation
Option 1: VSIX (Recommended)
- Download the latest
.vsixfrom Releases - Open VS Code / Cursor / Antigravity
Cmd+Shift+P→ Extensions: Install from VSIX → Select the file
Option 2: Build from Source
git clone https://github.com/wonseokjung/connect-ai.git
cd connect-ai
npm install
npm run compile
npx vsce package
⚙️ Engine Setup
Connect AI requires a local AI engine. Choose one:
Ollama (Recommended for Beginners)
# 1. Install
brew install ollama
# 2. Pull a model
ollama pull gemma3
# 3. Done — Ollama runs automatically in the background
LM Studio (Recommended for Apple Silicon)
- Download from lmstudio.ai
- Search and download a model (e.g., Gemma 4, Llama 3.1)
- Go to Developer tab (
<>icon) → Start Server - Ensure it shows
http://127.0.0.1:1234 - In Connect AI: ⚙️ Settings → Select LM Studio
💡 Tip: In LM Studio, set Context Length to 8192+ for best results.
Supported Models
| Model | Size | Best For |
|---|---|---|
| Gemma 4 E2B | 4.4 GB | Vision + Code (Recommended) |
| Gemma 3 | 3-5 GB | Fast general coding |
| Llama 3.1 | 4-8 GB | Multi-language support |
| Qwen 3 | 4-8 GB | Strong instruction following |
| DeepSeek Coder | 6-16 GB | Code-heavy tasks |
🧠 Second Brain (Knowledge Base)
Sync a GitHub repository as your personal knowledge base. Connect AI will reference it when answering questions.
- Click 🧠 button in the chat header
- Enter your GitHub repo URL
- Toggle knowledge mode ON
Your documents are stored locally at ~/.connect-ai-brain/.
🔒 Privacy
- Zero cloud dependency — No internet required after setup
- Zero data collection — All code stays on your machine
- Zero telemetry — No analytics, no tracking
Your code never leaves your computer. Period.
🛠️ Configuration
Access settings via the ⚙️ button in the chat panel:
| Setting | Default | Description |
|---|---|---|
| Engine | Ollama | Ollama or LM Studio |
| Temperature | 0.7 | Response creativity (0.0–1.0) |
| Top P | 0.9 | Nucleus sampling |
| Top K | 40 | Token selection range |
License
MIT — Free to use, modify, and distribute.
Designed & Developed by EZERAI × Connect AI