AgentOS Documentation

Learn how to build, register, and execute agents with AgentOS

Getting Started

1. Register Your Agent

Every agent must declare its capabilities and requirements to the AgentOS registry.

curl -X POST http://localhost:8000/agents/register \
  -H "Content-Type: application/json" \
  -d '{
    "name": "DataAnalyzer",
    "version": "1.0.0",
    "description": "Analyzes data and generates insights",
    "owner": "my-org",
    "capabilities": [
      {
        "name": "analyze",
        "type": "compute",
        "description": "Analyze input data",
        "input_schema": {...},
        "output_schema": {...}
      }
    ]
  }'

Key Concepts

Agent Registry

Global registry of all agents with their capabilities, versions, and performance metrics. Enables reliable agent-to-agent communication and discovery.

Capabilities

Structured declarations of what an agent can do. Each capability specifies input/output schemas, enabling type-safe agent composition.

Proof of Execution

Cryptographically signed records proving what an agent did, enabling radical transparency and auditability for compliance and debugging.

Composability

Agents call other agents with guaranteed semantics, automatic failure handling, and cross-agent transactions.

API Reference

POST /agents/register

Register a new agent in the registry

{
  "name": "string",
  "version": "string",
  "description": "string",
  "owner": "string",
  "capabilities": [Capability],
  "dependencies": [string]
}

GET /agents

List all registered agents with pagination

GET /agents/{agent_id}

Get detailed information about a specific agent

POST /agents/{agent_id}/execute

Execute an agent and generate proof of execution

GET /proofs/{execution_id}

Retrieve the proof of execution for an agent run

GET /stats

Get overall AgentOS statistics

Use Cases

Enterprise Agent Orchestration

Deploy multiple agents across the enterprise with complete visibility and control. AgentOS ensures agents work together reliably.

Compliance & Audit

Every agent action is cryptographically signed and auditable. Perfect for regulated industries requiring proof of execution.

Multi-Agent AI Systems

Build systems where agents reliably call other agents. AgentOS handles composition, failure recovery, and transaction semantics.

Agent Marketplace

Enable agents to be discovered and composed economically. AgentOS registry supports agent markets with billing and compensation.

Roadmap

✓

Phase 1: Agent Runtime & Registry

Standard agent interface, registry service, execution sandbox

→

Phase 2: Proof of Execution

Signed execution logs, audit trail, verification infrastructure

◯

Phase 3: Composability

Agent-to-agent calling, transaction support, failure handling

◯

Phase 4: Agent Economics

Resource pricing, compensation system, ROI tracking

◯

Phase 5: Advanced Monitoring

Anomaly detection, automatic recovery, performance optimization