demo-app

Demo App — Your First Governed Agent

A minimal example showing how to run an agent task through the Agent OS kernel with policy enforcement.

What It Does

Creates a StatelessKernel (no external dependencies), defines a simple agent function that processes a task through the kernel under a read_only policy, and prints the result. This is the simplest possible Agent OS example — one agent, one policy, one task.

The kernel intercepts the process_task action, checks it against the declared policy, and either allows or blocks execution. This demonstrates the core Agent OS pattern: all agent actions flow through the kernel, where governance policies are enforced before execution.

Prerequisites

• Python 3.10+
• Agent OS installed:

pip install -e "packages/agent-os[dev]"

How to Run

cd packages/agent-os/examples/demo-app
python agent.py

Expected Output

[Agent OS] Demo
========================================
[OK] Result: Processed: HELLO, AGENT OS!

Success! Your agent ran safely under kernel governance!

The kernel checked the 'read_only' policy before execution.

Next Steps

• Add more policies from packages/agent-os/examples/shared-policies/
• Try the self-evaluating example for agents that assess their own output quality
• See the governed-chatbot example for a conversational agent with full policy enforcement