The AI Agent Framework for Building, Deploying, and Orchestrating Intelligent Workflows

Create powerful AI agents • Design visual workflows • Extend with plugins • No code required

🌐 Website • ✨ Features • 📦 Installation • 🏃 Quick Start • 📚 API Docs • 🔌 Plugins • 💬 Discord

---

⭐ Star History

---

👥 Who Is This For?

AGNT is local-first and designed for:

• ✅ Single users - Run on your personal computer
• ✅ Families - Share across household devices
• ✅ Small teams - 2-10 people in your organization

NOT designed for:

• ❌ Multi-tenant SaaS (isolating hundreds of unrelated users)
• ❌ Public hosting (each org should self-host their own instance)
• ❌ Large enterprises (50+ concurrent users)

AGNT uses SQLite and real-time sync that broadcasts to all connected clients. Perfect for trusted groups sharing a workspace, not for isolating thousands of separate organizations.

---

🎯 Why AGNT?

🚀 Zero to Agent in Minutes No complex setup. No coding required. Just install, connect your AI provider, and start building intelligent agents. 🔗 Connect Everything 9+ AI providers, 50+ built-in tools, webhooks, and a plugin marketplace. If it has an API, AGNT can connect to it. 🎨 Visual-First Design Drag-and-drop workflow designer that makes complex automation accessible to everyone.

🔌 Infinitely Extensible Build custom plugins with our VSCode-style distribution system. Share with the community or keep them private. 🎯 Goal-Oriented AI Set objectives and let your agents work autonomously towards them with built-in progress tracking. 🖥️ True Desktop Power Native desktop app with full system access. No browser limitations. Your data stays on your machine.

---

✨ Features

🤖 AI Agent Management

 

• Multi-Provider Support — Connect to OpenAI, Anthropic, Google Gemini, Groq, Cerebras, DeepSeek, and more
• Custom Agents — Create specialized agents with unique personalities, capabilities, and tool access
• Agent Forge — AI-powered agent generation from natural language descriptions
• Real-time Chat — Interactive conversations with streaming responses
• Memory & Context — Agents remember conversation history and learn from interactions

🔧 Visual Workflow Designer

 

• Drag-and-Drop Interface — Build complex workflows visually without coding
• 50+ Built-in Tools — Triggers, actions, utilities, and widgets ready to use
• Conditional Logic — Branch workflows based on conditions and data
• Webhook Integration — Trigger workflows from external services
• Real-time Execution — Watch your workflows run with live status updates

🔌 Extensible Plugin System

┌─────────────────────────────────────────────────────────────┐
│  📦 Plugin Marketplace                                       │
├─────────────────────────────────────────────────────────────┤
│  🔷 Discord Bot        🔷 GitHub Actions    🔷 Slack         │
│  🔷 Gmail              🔷 Google Sheets     🔷 Notion        │
│  🔷 Twitter/X          🔷 Stripe            🔷 Zapier        │
│  🔷 YouTube            🔷 Dropbox           🔷 And more...   │
└─────────────────────────────────────────────────────────────┘

• VSCode/Figma-style Distribution — Install plugins as .agnt packages
• No Dependencies Required — Plugins come pre-bundled with all dependencies
• Hot Reload — Install and uninstall plugins without restarting
• Plugin Marketplace — Browse and install community plugins
• AI Plugin Generator — Describe what you need, AI builds the plugin

🎯 Goal-Oriented Automation

• Autonomous Goals — Set objectives and let agents work towards them
• Progress Tracking — Monitor goal completion and agent performance
• Golden Standards — Save successful executions as templates
• Evaluation Reports — AI-powered analysis of goal completion

🛠️ Developer Tools

• Tool Forge — Create custom tools with AI assistance
• MCP Server Support — Connect to Model Context Protocol servers
• Comprehensive API — Full REST API for programmatic control
• E2E Testing — Playwright-based test suite included
• WebSocket Streaming — Real-time updates and notifications

---

🖼️ Screenshots

---

📦 Installation

Prerequisites

Quick Install

# Clone the repository
git clone https://github.com/agnt-gg/agnt.git
cd agnt/desktop

# Install dependencies
npm install

# Install frontend dependencies
cd frontend && npm install && cd ..

# Start the application
npm start

🐳 Docker Install (Self-Hosting)

Use Docker for an isolated, production-ready deployment.

Option A: Pull Pre-built Images from GHCR (Recommended)

# Full version with browser automation (~1.5GB) - Port 33333
docker run -d \
  --name agnt-full \
  -p 33333:33333 \
  -v agnt-data:/root/.agnt/data \
  ghcr.io/agnt-gg/agnt:latest

# Access at http://localhost:33333

# Lite version without browser automation (~715MB) - Port 3333
docker run -d \
  --name agnt-lite \
  -p 3333:3333 \
  -v agnt-data:/root/.agnt/data \
  ghcr.io/agnt-gg/agnt:lite

# Access at http://localhost:3333

Available tags:

• latest / full - Latest Full variant with browser automation
• lite - Latest Lite variant without browser automation
• v0.3.7 / v0.3.7-full - Specific version (Full)
• v0.3.7-lite - Specific version (Lite)

Option B: Build from Source (Advanced)

# Clone the repository
git clone https://github.com/agnt-gg/agnt.git
cd agnt

# Option 1: Full version with browser automation (~1.5GB) - Port 33333
docker-compose up -d
# Access at http://localhost:33333

# Option 2: Lite version without browser automation (~715MB) - Port 3333
docker-compose -f docker-compose.lite.yml up -d
# Access at http://localhost:3333

# Option 3: Run both versions simultaneously
docker-compose -f docker-compose.both.yml up -d
# Full: http://localhost:33333
# Lite: http://localhost:3333

# Or use the Makefile
make run-both

Environment Configuration

Before running Docker, set up your environment file:

# Copy the example environment file
cp .env.example .env

Optional variables for Docker:

# Docker data directory (where persistent data is stored)
# Usually not needed - defaults to $HOME/.agnt/ which works on most systems
# Only set this if using Docker snap on Linux or want a custom location
# AGNT_HOME=/home/youruser

# App path for plugins (required for Docker)
APP_PATH=/app

Data Directory Structure:

Your data is stored at ~/.agnt/:

• SQLite database: ~/.agnt/data/agnt.db
• Plugins: ~/.agnt/data/plugins/
• Logs: ~/.agnt/logs/

Note for Docker snap users (Linux): If you installed Docker via snap, you must set AGNT_HOME with an absolute path (e.g., AGNT_HOME=/home/username docker-compose up -d) to avoid snap's home directory isolation issues.

Security secrets (change in production):

# Generate with: openssl rand -base64 32
JWT_SECRET=your-random-jwt-secret
SESSION_SECRET=your-random-session-secret
ENCRYPTION_KEY=your-random-encryption-key

See .env.example for all available configuration options including AI provider API keys, OAuth settings, and plugin configuration.

Docker variants available:

• 🔋 Full (~1.5GB): Includes Chromium for web scraping & browser automation (Port 33333)
• 🪶 Lite (~715MB): Smaller image without browser features, ~52% smaller (Port 3333)
• 🚀 Both: Run full and lite versions side-by-side for testing

Desktop installers available:

• 💻 Electron Full (~150-200MB): Portable desktop app with browser automation
• 📦 Electron Lite (~80-120MB): Lightweight desktop app, ~50% smaller

📖 See the Self-Hosting Guide for complete Docker setup, networking, and configuration.

📥 Download Pre-built Binaries

Platform-Specific Notes

🪟 Windows

Windows users may need to install build tools for native modules:

npm install --global windows-build-tools

🍎 macOS

Ensure Xcode Command Line Tools are installed:

xcode-select --install

🐧 GNU/Linux

Install required system dependencies:

# Ubuntu/Debian
sudo apt-get install build-essential libx11-dev libxkbfile-dev

# Fedora
sudo dnf install gcc-c++ make libX11-devel libxkbfile-devel

See GNU/Linux Build Instructions for detailed setup.

---

🏃 Quick Start

Development Mode

Run the frontend and backend separately for hot-reloading:

# Terminal 1: Start the frontend dev server
cd frontend
npm run dev

# Terminal 2: Start the Electron app
npm start

Production Mode

Build and run the production version:

# Build the frontend
cd frontend && npm run build && cd ..

# Start the app
npm start

---

🏗️ Building for Distribution

Create distributable packages for your platform:

# Build for current platform
npm run build

# Build for specific platforms
npm run build:win      # Windows (NSIS installer)
npm run build:mac      # macOS (DMG + ZIP)
npm run build:linux    # GNU/Linux (AppImage, DEB, RPM)

# Build for all platforms
npm run build:all

🪶 Lite Mode - Smaller Desktop Builds

Build AGNT Lite variants that exclude browser automation for ~50% smaller installers:

# Build Lite variant (~80-120MB vs ~150-200MB)
npm run build:lite
npm run build:lite:win      # Windows
npm run build:lite:mac      # macOS
npm run build:lite:linux    # GNU/Linux

# Build BOTH Full and Lite (recommended)
npm run build:both
npm run build:both:win
npm run build:both:mac
npm run build:both:linux

Lite Mode removes:

• ❌ Puppeteer/Playwright browser automation
• ❌ Web scraping tools
• ❌ Screenshot capture via browser

Everything else works:

• ✅ AI agents and workflows
• ✅ All API integrations
• ✅ Plugins, image processing, email automation

📖 See Electron Lite Mode Guide for details.

Build outputs are saved to the dist/ directory:

Build Artifacts

Platform	Full	Lite	Size Reduction
Windows	AGNT-0.3.7-win-x64.exe (~150MB)	AGNT-Lite-0.3.7-win-x64.exe (~80MB)	~47%
macOS	AGNT-0.3.7-mac-x64.dmg (~200MB)	AGNT-Lite-0.3.7-mac-x64.dmg (~120MB)	~40%
GNU/Linux	AppImage (~180MB), DEB, RPM	AppImage (~100MB), DEB, RPM	~44%

---

🔌 Plugin Development

AGNT features a powerful plugin system that allows you to extend functionality with custom tools.

Creating a Plugin

# Navigate to the plugins development directory
cd backend/plugins/dev

# Create your plugin directory
mkdir my-awesome-plugin
cd my-awesome-plugin

Create a manifest.json:

{
  "name": "my-awesome-plugin",
  "version": "1.0.0",
  "description": "Does something awesome",
  "author": "Your Name",
  "tools": [
    {
      "type": "awesome-tool",
      "entryPoint": "./awesome-tool.js",
      "schema": {
        "title": "Awesome Tool",
        "category": "action",
        "description": "Performs awesome operations",
        "parameters": {
          "input": {
            "type": "string",
            "description": "Input value"
          }
        }
      }
    }
  ]
}

Create your tool implementation (awesome-tool.js):

class AwesomeTool {
  constructor() {
    this.name = 'awesome-tool';
  }

  async execute(params, inputData, workflowEngine) {
    // Your tool logic here
    return {
      success: true,
      result: `Processed: ${params.input}`,
      error: null,
    };
  }
}

export default new AwesomeTool();

Build and install:

cd backend/plugins
node build-plugin.js my-awesome-plugin

📚 See the full Plugin Development Guide for more details.

---

🗂️ Project Structure

desktop/
├── 📄 main.js              # Electron main process
├── 📄 preload.js           # Preload scripts for IPC
├── 📄 package.json         # Project configuration
│
├── 📁 backend/             # Express.js backend server
│   ├── 📄 server.js        # Server entry point
│   ├── 📁 src/
│   │   ├── 📁 routes/      # API route handlers
│   │   ├── 📁 services/    # Business logic
│   │   ├── 📁 models/      # Data models
│   │   └── 📁 plugins/     # Plugin management
│   └── 📁 plugins/         # Plugin development & builds
│
├── 📁 frontend/            # Vue.js frontend application
│   ├── 📁 src/
│   │   ├── 📁 views/       # Page components
│   │   ├── 📁 components/  # Reusable components
│   │   ├── 📁 store/       # Vuex state management
│   │   └── 📁 services/    # API services
│   └── 📁 public/          # Static assets
│
├── 📁 build/               # Build resources & icons
├── 📁 docs/                # Documentation
└── 📁 tests/               # E2E test suites

---

🧪 Testing

Run the end-to-end test suite:

# Run all tests
npm run test:e2e

# Run specific test file
npx playwright test tests/e2e/agents.spec.js

See Testing Instructions for more details.

---

📖 Documentation

Document	Description
📚 API Documentation	Complete REST API reference
🔨 Build Instructions	Detailed build guide
🐧 GNU/Linux Build Guide	GNU/Linux-specific setup
🐳 Self-Hosting Guide	Docker deployment & hosting
🪶 Docker Lite Mode	Docker without browser automation
🪶 Electron Lite Mode	Smaller desktop builds (~50% size)
🔌 Plugin Development	Creating custom plugins
🔧 Rebuild Guide	Native module rebuilding
🚀 CI/CD Pipelines	GitHub Actions workflows

---

🔗 Supported AI Providers

Provider	Models	Status
OpenAI	GPT-4, GPT-4 Turbo, GPT-3.5	✅ Supported
Anthropic	Claude 3.5, Claude 3	✅ Supported
Google	Gemini Pro, Gemini Ultra	✅ Supported
Groq	Llama 3, Mixtral	✅ Supported
Cerebras	Fast inference models	✅ Supported
DeepSeek	DeepSeek Coder, Chat	✅ Supported
OpenRouter	100+ models	✅ Supported
Together AI	Open source models	✅ Supported
Custom	Any OpenAI-compatible API	✅ Supported

---

🤝 Contributing

We welcome contributions! Here's how you can help:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Development Guidelines

• Follow the existing code style
• Write tests for new features
• Update documentation as needed
• Keep commits atomic and well-described

---

📊 Project Stats

---

📄 License

This project is licensed under a Custom License. See LICENSE.md for details.

---

🙏 Acknowledgments

• Built with Electron
• Frontend powered by Vue.js

---

Built with ❤️ by Nathan Wilbanks

_{If you find AGNT useful, please consider giving it a ⭐ on GitHub!}