Getting Started with OnPremiseAgent in Under 10 Minutes

OnPremiseAgent is designed so that a single engineer can go from zero to a running AI agent in under 10 minutes. No cloud accounts to configure, no third-party API keys to manage, no data leaving your network.

This guide walks you through the complete setup.

Prerequisites

Before you start, make sure you have:

• Docker (20.10+) and Docker Compose (v2) installed
• Node.js 20+ for the OPA CLI
• 8GB RAM minimum (16GB recommended for production)
• A valid OnPremiseAgent license (start a free trial at onpremiseagent.com/pricing)

Step 1: Install the OPA CLI

The OnPremiseAgent CLI is the primary tool for managing agents locally. Install it globally via npm:

npm install -g @onpremiseagent/cli

Verify the installation:

opa --version
# OnPremiseAgent CLI v1.0.0

Step 2: Authenticate

Log in with your license credentials:

opa login

This stores your license token locally and caches the public key for offline validation. The license daemon will send only heartbeat signals (no customer data) to verify your subscription status.

Step 3: Pull an Agent Template

OnPremiseAgent ships with 6 pre-built agent templates optimized for common enterprise use cases:

Pull your chosen template:

opa pull rag-knowledge

This downloads the agent's Docker Compose stack, SKILL.md files, and default configuration to your local .opa/ directory.

Step 4: Configure and Start

Launch the agent with a single command:

opa start --name my-first-agent \
  --template rag-knowledge \
  --provider anthropic \
  --model "claude-sonnet-4-5-20250514" \
  --lang en

Behind the scenes, this:

1. Generates a Docker Compose file with the agent runtime, license daemon, and any required services (e.g., ChromaDB for RAG)
2. Validates your license locally using the cached public key
3. Starts the agent stack
4. Binds the license to your hardware fingerprint

Step 5: Verify

Check that your agent is running:

opa status

You should see output like:

Agent: my-first-agent
Template: rag-knowledge
Status: running
Uptime: 45s
License: valid (expires 2027-03-26)
Channels: slack

What Happens Under the Hood

The agent stack consists of three containers:

1. opa-runtime — The Node.js agent execution engine that processes queries, invokes tools, and generates responses. Skills run in sandboxed environments with no filesystem or network access beyond declared tools.

1. opa-license-daemon — A lightweight process (<50MB memory) that validates your license locally and sends a heartbeat to the license server every 24 hours. If the internet is unavailable, the agent continues running for up to 30 days (grace period).

1. Service containers — Template-specific services like ChromaDB (for RAG vector storage) or Redis (for caching).

Next Steps

• Add documents to your RAG agent: opa ingest --name my-first-agent ./documents/
• Connect channels like Slack or Teams: opa config set --name my-first-agent --channel slack
• Enable security features: output filtering, audit logging, and skill allowlists
• Deploy to production with Kubernetes using the Helm chart: opa start --method kubernetes

For production deployments with custom skills, compliance configuration, and multi-agent workflows, schedule a demo with our team.

Need help? Email us at contact@onpremiseagent.com or explore the documentation.