How to Deploy Agent-Memo on Your Own Server with Docker

TL;DR: Deploy Agent-Memo on your own server in under 10 minutes. One install script sets up Docker, PostgreSQL, the embedding model, and the admin dashboard. No external dependencies after initial setup.

Prerequisites

• A Linux server (Ubuntu 22.04 or 24.04 recommended)
• Minimum 4 CPU cores and 8GB RAM (the BGE-M3 embedding model needs ~2GB)
• Docker and Docker Compose installed
• A valid Agent-Memo self-hosted license key

Step 1: Download and install

Download the release tarball and run the install script:

tar -xzf agent-memo-private-v1.0.0.tar.gz
cd agent-memo-private
chmod +x install.sh
./install.sh

The install script generates secure random passwords for PostgreSQL and JWT, creates the Docker Compose configuration, and builds the container image.

Step 2: Start the services

docker compose up -d

This starts PostgreSQL and the Agent-Memo application. The first boot takes 1-2 minutes as the embedding model initializes. Subsequent starts are fast.

Step 3: Run the setup wizard

Open http://your-server:8080 in your browser. The setup wizard guides you through 4 steps:

1. License — Paste your license key for validation
2. Admin account — Create the first admin user
3. Configuration — Set company name and preferences
4. Done — Login and start using the dashboard

Step 4: Connect your agents

Create .mcp.json in your project root:

{
  "mcpServers": {
    "agent-memo": {
      "command": "npx",
      "args": ["-y", "@agent-memo/mcp-client"],
      "env": {
        "AGENTMEMO_TOKEN": "your-api-token",
        "AGENTMEMO_API_URL": "http://your-server:8080"
      }
    }
  }
}

Get your API token from the dashboard (Settings → API tab). The token is a standard JWT — same format as the cloud version.

Architecture

The self-hosted deployment is a single Docker Compose stack:

• app container — Node.js backend + Next.js frontend + BGE-M3 embedding model (~1GB image)
• postgres container — PostgreSQL 16 with pgvector for semantic search

Data is persisted in Docker volumes. The embedding model is bundled in the image, so no internet access is needed after deployment.

Backup and restore

# Backup
./backup.sh    # Creates timestamped SQL dump

# Restore
./restore.sh backup-2026-04-10.sql

Upgrading

# Download new release, then:
./upgrade.sh

The upgrade script rebuilds the Docker image and restarts the container. Database migrations run automatically on startup.

SSL/HTTPS

For production use, put a reverse proxy (Nginx, Caddy, or Traefik) in front of the application with SSL termination. The app listens on port 8080 by default.

Troubleshooting

• Container won't start — Check docker compose logs app. Most common: insufficient RAM for the embedding model
• Slow first request — The embedding model loads on first API call (~10 seconds). Subsequent calls are fast
• Agent can't connect — Ensure the server is reachable from the developer's machine. Check firewall rules for port 8080

Get a license key | View pricing