Introduction

AI agents emerged as an offshoot of the ongoing AI wave, which has evolved from GPTs (Generative Pre-trained Transformers), passive chat-based models, to autonomous entities capable of executing tasks without human intervention. This shift, also dubbed AI 2.0, led to the integration of AI agents into various Web3 functions. We have seen AI agents create wallet addresses, manage social media accounts, and even launch their own tokens.

One innovative product from this narrative was ElizaOS, a decentralized venture capital DAO run by AI agents. Originally launched as AI16z in October 2024 as a play on the well-known venture capital firm a16z, the project was later rebranded to ElizaOS to avoid confusion. However, its token ticker remains $ai16z (as of the time of writing). The first version of ElizaOS also served as a toolbox for developers to build and customize AI agents.

With ElizaOS v2, the project leaps forward, addressing the technical issues that affected user and developer experience, introducing new products, and launching a $10 million fund dedicated to the development of open-source AI.

What is ElizaOS? - A Recap of v1

Source: Eliza GitHub

ElizaOS is an open-source operating system designed for developing, deploying, and managing autonomous AI agents. Built with the high-level programming language TypeScript, ElizaOS is a flexible and extensible platform that allows developers to customize and expand its functionality to suit specific requirements.

Features of ElizaOS v1

The first iteration of ElizaOS found rapid success with its ability to build an AI agent. Within six months of release, it had 4.8K forks of its core repository, and 500+ contributors advancing the open-source ecosystem with over 100 plugins. Some of its features include;

Social Media Integration

ElizaOS could deploy multiple agents each with their own persona and function on Discord, X, Telegram, and more.

Web3 Compatible

The framework was designed to interact with blockchain systems, allowing AI agents to perform tasks such as token transactions and smart contract interactions.

Media Processing

ElizaOS could analyze and interact with various data formats, from PDFs to audio files.

Problems with ElizaOS v1

Overcrowded Core Codebase

The core of ElizaOS v1 contained an excessive number of packages, leading to a cluttered and less sustainable architecture. This made it difficult for developers to navigate the code and add new features without causing issues.

Limited Message Routing Between Platforms

ElizaOS v1 struggled to allow AI agents to talk to each other across different platforms like Twitter and Discord. Agents were isolated and unable to share information or work together.

Fragmented Wallet Management

The system used separate wallets for different blockchain networks, thus, managing multiple wallets was cumbersome and increased the chances of errors, especially when agents needed to handle transactions across various blockchains.

Rigid State Composition and Limited Planning

ElizaOS v1 had hard-coded state compositions, restricting the flexibility and autonomy of AI Agents as they had limited ability to plan for complex tasks.

Technical Glitches

Users reported various technical issues, such as errors when starting the agent or issues with package installations. These challenges affected the overall user experience and system reliability.

ElizaOS vs Virtuals Protocol - Developer Ecosystem Comparison

Virtuals Protocol is a platform for creating, tokenizing, and co-owning autonomous AI agents, providing tools for deployment and customization to suit various applications and project needs.

Here’s a comparative analysis of the developer ecosystems for ElizaOS V2 and Virtuals Protocol, based on their GitHub repositories.

ElizaOS v2

Source: ElizaOS

ElizaOS v2 was first announced at the Catstanbul, an event organized by Jupiter in January 2025.

During the event, ElizaOS founder Shaw (@shawmakesmagic) highlighted the problems with ElizaOS v1 and announced the development of a v2 to upgrade the Eliza Architecture.

The beta version of ElizaOS V2 was launched on March 18, 2025, and the full product launch is scheduled for April 2025.

Improvements in ElizaOS v2

Modular Framework

ElizaOS v2 uses a Package registry system and a CLI (command-line interface) that makes it easier for developers to customize and add new features without modifying the core code.

Unified Abstraction Layer

ElizaOS v2 introduces a system that allows AI agents to handle assets across multiple blockchain networks through a unified wallet system.

Improved Task Management

ElizaOS v2 adopts an event-driven architecture that allows AI agents to respond in real time to data updates.

It also uses Hierarchical Task Networks that allow AI agents to break down complex tasks into structured steps and dynamically adjust plans as environments change.

AI Autonomy

AI agents in ElizaOS v2 can go beyond simple commands and independently manage workflows, run businesses, and develop financial strategies.

Upcoming ElizaOS Products

Aside from upgrading the Eliza framework, the team plans to float Eliza Labs as the research and development engine for decentralized AI projects with real-world applications. It plans to pioneer new agent-based projects while supporting open-source contributors via grants, accelerator programs, and ecosystem funding.

Some of the major initiatives currently in development include:

The Agent Marketplace

The Agent Marketplace is a token launchpad and a no-code platform for building simple AI agents. It introduces multi-agent functionality, allowing multiple autonomous agents to interact and collaborate within a decentralized framework. AI-enhanced tools simplify token creation and agent deployment for developers and non-technical users.

Eliza Studios

Eliza Studios is a creative studio where Artificial Intelligence powers art, storytelling, and digital experiences. The team plans to build autonomous characters, generative media experiments, and immersive experiences that will redefine entertainment.

DegenSpartanAI

DegenSpartanAI is a crypto-native AI trading agent. It interacts across social platforms like X, Discord, and Telegram, engaging in conversations without human intervention.

The ElizaOS team plans to evolve the Agent from trading strategies and its comedic commentary, it will evolve into an interactive AI with real-time market insights, user-driven discussions, and NFT collaborations.

In the long term, it aims to become a fully autonomous trading agent, integrating multi-platform execution, adaptive learning, and a verifiable track record within the Global Trust Marketplace.

The Autonomous Investor

An intelligent investment platform that integrates social trading, reputation scoring, and decentralized execution. Users can submit trade suggestions, which are evaluated for credibility through the Trust Marketplace, and the platform aggregates data from multiple social platforms.

In the long term, the project will evolve into a multi-instance system, enabling groups to define their own risk parameters, manage DAO treasuries, and use customizable AI modules to enhance investment strategies.

Block Tank

Block Tank is an AI-driven game show where participants pitch investment ideas in an interactive environment.

Users submit a pitch with details about their personality and project idea, and a group of AI judges accordingly.

The popular business TV Series “Shark Tank” inspires the game.

ElizaOS and Jupiter Magic Fund

The Magic Fund program is a $10 million fund launched by ElizaOS in partnership with Jupiter to support open-source AI development in Web3.

Risks Associated with AI Agents in Web3

AI agents in Web3 introduce various risks that stem from their autonomous decision-making, interaction with blockchain protocols, and handling of financial transactions. Some key risks and recommended solutions include:

Security Risks

Integrating AI into Web3 amplifies security risks, particularly in how these agents interact with smart contracts, process data, and execute transactions. If an AI agent relies on natural language inputs (e.g., for trading or governance), attackers can manipulate responses to alter AI behavior, Oracle dependencies also pose a risk. AI agents rely on external data sources, and if these sources are compromised, the AI may act on false information. A manipulated oracle could, for instance, trick an AI-driven trading bot into panic-selling assets at a loss. Similarly, smart contract vulnerabilities like reentrancy attacks become more dangerous when AI agents interact autonomously, as they might unknowingly trigger unintended transaction loops.

Recommended safeguards include adversarial training, secure model deployment through trusted execution environments (TEEs), and multi-oracle verification to ensure reliable data inputs. Limiting an AI agent’s permissions and integrating circuit breakers can prevent it from executing damaging actions.

Regulatory Risks

AI in Web3 operates across decentralized, often unregulated environments, which can lead to legal challenges. AI agents executing transactions could unknowingly facilitate money laundering or interact with sanctioned addresses, exposing platforms to regulatory scrutiny. Liability is another gray area. If an AI agent makes an unauthorized trade or a damaging financial decision, who is responsible? The developer, the platform deploying the AI, or the users relying on it? Concerns regarding user privacy could also arise.

As of this writing, the only major regulatory framework for artificial intelligence is the European Union’s Artificial Intelligence Act. It identifies unacceptable uses of AI as AI systems that manipulate human behavior against their free will, and Social scoring systems based on their social behavior.

It classifies AI systems based on risk levels, imposing strict requirements on high-risk applications. It enforces high-risk AI systems to establish a continuous risk management system, regularly reviewed and updated throughout their lifecycle. This includes identifying and analyzing potential risks to health, safety, or fundamental rights, and adopting measures to manage them.

Conclusion

ElizaOS v2 redefines AI deployment in the Web3 ecosystem, addressing technical flaws found in its earlier version while expanding its capabilities with a modular architecture and hierarchical task networks. This latest version provides a range of new features that allow AI agents to perform complex tasks with autonomy, improve user experience, and manage investments.

The team behind ElizaOS is also committed to driving decentralized AI, backing it with a $10 million fund to support open-source AI development in Web3.

ElizaOS will lay the groundwork for a new phase of the AI-powered future, serving as a vital resource for AI applications in the Web3 space.

You can check out the latest updates on ElizaOS on their official website, GitHub, and X.