A lightweight, lightning-fast, and highly secure Personal AI Agent running locally on your hardware. Inspired by OpenClaw and designed as its ultra-lightweight, zero-bloat alternative. Support for both local models (Ollama, LM Studio) and cloud APIs (OpenRouter, Gemini, OpenAI). Avoid bloated frameworks that devour your RAM. Just clean TypeScript, modular markdown prompt design, and absolute data privacy.
Beautiful cyberpunk-themed local Web UI for focused work, plus a remote Telegram bot to command your agent securely on the go.
Attach images directly to the chat. The agent automatically resizes and optimizes files locally before sending them to the LLM.
Customize personality, identity, and behavior on the fly. Instructions are compiled dynamically from simple, git-ignored Markdown files.
No manual editing of .env. Local setup wizard configures everything (including local LLM endpoints or cloud APIs) and writes to a git-ignored JSON config.
File system manager (restricted to a dedicated workspace folder), real-time web scraper, and image generation with automatic API provider fallback.
Powered by SQLite and the ultra-lightweight sqlite-vec extension. Full session
storage and semantic vector similarity search.
Node.js, Fastify (high speed, ultra-low overhead), TypeScript.
EJS templates, vanilla CSS (Cyberpunk theme), reactive vanilla JS.
better-sqlite3 for configuration and logs + sqlite-vec for local vector
embeddings.
sharp for blazing-fast local image resizing and optimization.
axios for connections and cheerio for clean HTML processing.
telegraf integration for secure connection with the Telegram Bot API.
Get your local agent up and running in less than a minute. No complex databases, Docker, or external dependencies required.
Ensure you have Node.js (v20.x or later) and npm (v9.x or later) installed on your system.
Run these commands to clone the repository and download dependencies:
git clone https://github.com/nordevelopment/OpenPersonalAIAgent.git
cd OpenPersonalAIAgent
npm installStart the application. It will automatically build and start the local server. On the first launch, it will redirect you to the visual web settings panel to set your API keys:
npm startOpen http://127.0.0.1:3000 in your browser. (For active
development with hot-reload, run npm run dev instead).
If you prefer using traditional environment variables, copy the template and edit it manually:
cp .env.example .envCompare PAIAgent with heavyweight alternatives. The metrics prove how lightweight and efficient it is for local hardware.
| Metric / Feature | 🤖 PAIAgent | 🐳 Open WebUI | 💬 LibreChat | 📦 AnythingLLM | 🦞 OpenClaw |
|---|---|---|---|---|---|
| RAM Footprint | ~50 - 100 MB | 1.5 GB+ (Docker) | 1 GB+ (Docker Compose) | 500 MB+ (Electron/Docker) | 200 MB+ (pnpm/Docker) |
| Startup Time | < 1 second | 30 - 60 seconds | 30 - 60 seconds | 10 - 20 seconds | 5 - 10 seconds |
| Installation | < 1 minute (One-command) | 5+ minutes (Docker setup) | 10+ minutes (Complex Compose) | 5+ minutes (Docker / Installer) | 5+ minutes (pnpm / Docker) |
| Dependencies | Node.js & npm (Native) | Docker & Ollama | Docker, Mongo, Redis | Docker or Desktop | Node.js & pnpm (or Docker) |
| Database | SQLite (embedded) | PostgreSQL / MySQL | MongoDB + Meilisearch | Vector DB + SQLite | SQLite (embedded) |
| Ideal For | Lightweight personal workspace | Multi-user hosting | Enterprise chat solutions | Document-based RAG | Multi-channel agent bots |
PAIAgent is open-source and free. If it saved your hardware resources and simplified your AI operations, consider supporting our development node.
[ SPONSOR ON GITHUB ]