Two images, one decision: full ships everything; lite is the smaller footprint for constrained hosts. Both run the same runtime your desktop app runs.

Full image

Get AGNT running with full browser automation in under 5 minutes.

What You Get

• ✅ Complete web application (browser-based UI)
• ✅ Browser automation (Puppeteer/Playwright)
• ✅ Multi-device access
• ✅ Supports 2-10 concurrent users
• 📦 Image size: ~1.5GB
• 🌐 Port: 3333

Prerequisites

• Docker 20.10+ or Docker Desktop (any OS: Windows, macOS, GNU/Linux, FreeBSD, etc.)
• 2GB free RAM
• 2GB free disk space

Installation

Option 1: Pull from GHCR (Recommended - Fastest)

# Run container with pre-built image
docker run -d \
  --name agnt-full \
  -p 3333:3333 \
  -v agnt-data:/app/data \
  -e NODE_ENV=production \
  -e BASE_URL=http://localhost:3333 \
  --restart unless-stopped \
  ghcr.io/agnt-gg/agnt:latest

Using specific version:

docker run -d \
  --name agnt-full \
  -p 3333:3333 \
  -v agnt-data:/app/data \
  -e NODE_ENV=production \
  -e BASE_URL=http://localhost:3333 \
  --restart unless-stopped \
  ghcr.io/agnt-gg/agnt:v0.5.0

Option 2: Docker Compose with GHCR

Create docker-compose.yml:

version: '3.8'

services:
  agnt-full:
    image: ghcr.io/agnt-gg/agnt:latest
    container_name: agnt-full
    ports:
      - "3333:3333"
    environment:
      - NODE_ENV=production
      - BASE_URL=http://localhost:3333
    volumes:
      - agnt-data:/app/data
    restart: unless-stopped

volumes:
  agnt-data:

Start the service:

docker-compose up -d

Option 3: Build from Source (Advanced)

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

# Start Docker Full
make run-full

# Or use docker-compose directly
docker-compose up -d

Access AGNT

Open your browser to:

http://localhost:3333

From other devices on your network:

http://YOUR_SERVER_IP:3333

Verify Installation

# Check container status
docker ps | grep agnt

# View logs
docker logs agnt -f

Healthy status should show:

STATUS: Up X seconds (healthy)

First Steps

1. Open web UI: http://localhost:3333
2. Configure AI provider: Add your OpenAI/Anthropic/Google API key
3. Create first agent: Click "New Agent" and configure
4. Test chat: Send a message to your agent
5. Explore workflows: Create automated workflows

Common Commands

# Stop AGNT
docker stop agnt

# Start AGNT
docker start agnt

# View logs
docker logs agnt -f

# Restart AGNT
docker restart agnt

# Remove AGNT (keeps data)
docker rm agnt

# Update to latest version
docker pull ghcr.io/agnt-gg/agnt:latest
docker stop agnt
docker rm agnt
# Re-run docker run command above

Data Location

All data stored in:

• Database: ~/.agnt/data/agnt.db
• Plugins: ~/.agnt/plugins/installed/
• Logs: ~/.agnt/logs/

Features Available

✅ All AI agents and workflows
✅ All API integrations (OpenAI, Anthropic, Google, etc.)
✅ Plugin system
✅ Web scraping with Puppeteer
✅ Browser automation with Playwright
✅ Screenshot capture
✅ Image processing
✅ Email automation
✅ MCP server support

Troubleshooting

Port 3333 already in use

Change port in docker run command:

-p 3334:3333  # Use port 3334 instead

Container won't start

Check logs:

docker logs agnt

Can't access from other devices

Ensure firewall allows port 3333:

# Ubuntu/Debian
sudo ufw allow 3333

# Check if port is open
netstat -tlnp | grep 3333

Next Steps

• README - Learn how to use AGNT
• Plugin Development - Create custom plugins
• API Documentation - REST API reference
• Switch to Lite Mode - Smaller image without browser

Support

• Issues: GitHub Issues
• Discord: Join community
• Documentation: Full docs

Lite image

Get AGNT running lightweight (no browser automation) in under 5 minutes.

What You Get

• ✅ Complete web application (browser-based UI)
• ✅ All core features (AI agents, workflows, plugins)
• ✅ Multi-device access
• ✅ Supports 2-10 concurrent users
• ❌ No browser automation (Puppeteer/Playwright)
• 📦 Image size: ~715MB (52% smaller than Full)
• 🌐 Port: 3333

Prerequisites

• Docker 20.10+ or Docker Desktop (any OS: Windows, macOS, GNU/Linux, FreeBSD, etc.)
• 1GB free RAM
• 1GB free disk space

Installation

Option 1: Pull from GHCR (Recommended - Fastest)

# Run container with pre-built image
docker run -d \
  --name agnt-lite \
  -p 3333:3333 \
  -v agnt-data:/app/data \
  -e NODE_ENV=production \
  -e BASE_URL=http://localhost:3333 \
  --restart unless-stopped \
  ghcr.io/agnt-gg/agnt:lite

Using specific version:

docker run -d \
  --name agnt-lite \
  -p 3333:3333 \
  -v agnt-data:/app/data \
  -e NODE_ENV=production \
  -e BASE_URL=http://localhost:3333 \
  --restart unless-stopped \
  ghcr.io/agnt-gg/agnt:v0.5.0-lite

Option 2: Docker Compose with GHCR

Create docker-compose.yml:

version: '3.8'

services:
  agnt-lite:
    image: ghcr.io/agnt-gg/agnt:lite
    container_name: agnt-lite
    ports:
      - "3333:3333"
    environment:
      - NODE_ENV=production
      - BASE_URL=http://localhost:3333
    volumes:
      - agnt-data:/app/data
    restart: unless-stopped

volumes:
  agnt-data:

Start the service:

docker-compose up -d

Option 3: Build from Source (Advanced)

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

# Start Docker Lite
make run-lite

# Or use docker-compose directly
docker-compose -f docker-compose.lite.yml up -d

Access AGNT

Open your browser to:

http://localhost:3333

From other devices on your network:

http://YOUR_SERVER_IP:3333

Verify Installation

# Check container status
docker ps | grep agnt-lite

# View logs
docker logs agnt-lite -f

Healthy status should show:

STATUS: Up X seconds (healthy)

First Steps

1. Open web UI: http://localhost:3333
2. Configure AI provider: Add your OpenAI/Anthropic/Google API key
3. Create first agent: Click "New Agent" and configure
4. Test chat: Send a message to your agent
5. Explore workflows: Create automated workflows

Common Commands

# Stop AGNT
docker stop agnt-lite

# Start AGNT
docker start agnt-lite

# View logs
docker logs agnt-lite -f

# Restart AGNT
docker restart agnt-lite

# Remove AGNT (keeps data)
docker rm agnt-lite

# Update to latest version
docker pull ghcr.io/agnt-gg/agnt:lite
docker stop agnt-lite
docker rm agnt-lite
# Re-run docker run command above

Data Location

All data stored in:

• Database: ~/.agnt/data/agnt.db
• Plugins: ~/.agnt/plugins/installed/
• Logs: ~/.agnt/logs/

Features Available

✅ All AI agents and workflows
✅ All API integrations (OpenAI, Anthropic, Google, etc.)
✅ Plugin system
✅ Image processing (non-browser)
✅ Email automation
✅ MCP server support
✅ PDF reading
✅ Data transformations

Features NOT Available

❌ Web scraping with Puppeteer
❌ Browser automation with Playwright
❌ Screenshot capture via browser
❌ HTML to PDF via browser

Need these features? Use Docker Full instead.

Troubleshooting

Port 3333 already in use

Change port in docker run command:

-p 3334:3333  # Use port 3334 instead

Container won't start

Check logs:

docker logs agnt-lite

Browser automation tool fails

This is expected - Lite mode doesn't include browser automation.
Tools requiring Puppeteer/Playwright will show error:

Browser automation is not available in AGNT Lite Mode

Switch to Docker Full if you need browser features.

Can't access from other devices

Ensure firewall allows port 3333:

# Ubuntu/Debian
sudo ufw allow 3333

# Check if port is open
netstat -tlnp | grep 3333

Upgrading to Full

Want browser automation features later?

# Stop lite container
docker stop agnt-lite
docker rm agnt-lite

# Start full container (uses same data)
docker-compose up -d

# Access on new port
http://localhost:3333

Your data is preserved - just change the Docker image!

Next Steps

• README - Learn how to use AGNT
• Plugin Development - Create custom plugins
• API Documentation - REST API reference
• Upgrade to Full - Get browser automation

Support

• Issues: GitHub Issues
• Discord: Join community
• Documentation: Full docs