Agents Architecture

What Are Agents?

An agent is a specialized process with defined capabilities and objectives. Agents can perform simple tasks, such as fetching data, or operate autonomously by executing complex logic, including multi-step workflows, conditional decision-making, and dynamic interactions with other agents.

Key Components:

Agents are composed of one or more AO processes and may integrate other network features as part of their configuration.

Processes

Processes are decentralized, deterministic computing environments that interact with the AO network based on predefined logic. They function similarly to smart contracts but offer greater flexibility and access to unbounded compute. Each process maintains a persistent, holographic state, stored as a log of messages on Arweave, which can be accessed for verification. These features enable AO processes to function as verifiable autonomous agents, capable of performing large-scale computations onchain and executing advanced tasks without human intervention.

A unique feature of AO processes is that they can natively run large language models (LLMs) within the network. For more details, refer to the "AI on AO" section.

Cron Messages

Cron messages are a feature of AO's architecture, allowing processes to be triggered at specific intervals or in response to network events. Agents can integrate cron messaging to perform recurring tasks, such as streaming data or maintaining platform services, without relying on off-chain processing or external triggers.

Autonomous Agents

Autonomous agents consist of one or more processes that execute tasks autonomously without human intervention. They are capable of dynamically responding to network events, cron messages, and data signals, and can access LLM processes as well as spawn additional agents to increase capacity.

Autonomous Agent Capabilities:

• Operate 24/7 without human oversight

• Perform complex, multi-step tasks autonomously

• Coordinate with other autonomous agents

• Build and scale platforms, applications, and services

• Access and respond to real-time data signals

Platforms and Applications

In the AO ecosystem, platforms and applications are composed of networks of interconnected autonomous agents that coordinate to perform tasks, manage workflows, and execute decentralized services.

Platforms and applications make it possible for networks of agents to perform tasks, manage workflows and execute services like trading strategies as a decentralized dynamic system. For the human users of our platforms and applications, agents may play the role of trading counterparties, but also of employable, customizable service providers.

Check out some Agents listed on Dexi.