Protocol

Overview

The AOS Protocol provides a standardized message structure that enables agents to exchange intents, capabilities, actions, and environment data.

This section outlines the core schema, validation rules, and examples. You can expand each block later with full documentation.

Core Schema

{
  "version": "1.0",
  "agent_id": "your-agent-id",
  "capabilities": ["search", "analysis", "execution"],
  "endpoints": {
    "primary": "/api/agent"
  }
}

At minimum, an AOS node must expose a version identifier, an agent ID, and one or more supported capabilities.

Capabilities

Capabilities describe what actions an agent can perform. These can range from basic tasks like text analysisto more advanced features like autonomous task planning.

• search – Query external sources or internal datasets.
• analysis – Perform structured or unstructured reasoning.
• execution – Execute actions or workflows programmatically.

Endpoints

Agents must expose a primary endpoint where all protocol-compliant requests are sent. This may be a serverless function, a REST route, or an edge handler.

POST /api/agent

{
  "intent": "search",
  "query": "latest agent frameworks"
}

Validation

Before an agent processes a request, it must validate structure and intent fields. Invalid messages should respond with meaningful error contexts.