OKX Ventures Research Report: Analyzing 10+ Projects to Help You Understand the AI Agent Landscape (Part 2)

To better achieve value capture, we will evaluate projects based on the following framework, covering multiple assessment items such as whether to open source, key differentiating factors from existing AI protocols, long-term revenue channels, and ecosystem transaction volumes.

1. DeFAI

DeF AI combines the advantages of DeFi and AI, aiming to simplify the complex operations of DeFi, making it easy for ordinary users to use these financial tools. With the introduction of AI technology, DeF AI can automate complex financial decisions and trading processes, lowering the technical threshold for users while enhancing operational efficiency and intelligence. Although the current market size of DeF AI is less than $1 billion, far below the $100 billion DeFi market, this also means that DeF AI has enormous growth potential.

Griffain: AI Application Store in the Solana Ecosystem

Griffain is an AI agent engine built on the Solana blockchain, designed to simplify cryptocurrency operations through natural language interaction, integrating core functions such as wallet management, token trading, NFT minting, and DeFi strategy execution. The project was founded by Tony Plasencia, initially proposed at a Solana hackathon, and received support from Solana founder Anatoly Yakovenko. As the first high-performance abstract AI agent in the Solana ecosystem, Griffain combines natural language processing (NLP) technology to provide a user experience similar to Copilot and Perplexity, driving the evolution of AI-driven on-chain interaction models.

Griffain uses Shamir Secret Sharing (SSS) technology to split wallet keys, ensuring the security of user assets. Core features include natural language trading commands (supporting DCA, limit orders, etc.), AI agent collaborative task execution, market analysis (data parsing such as position distribution), and integrated token issuance and NFT minting on the pumpfun platform. At the same time, the platform offers personalized AI agents, allowing users to adjust commands based on their needs to execute on-chain tasks; special AI agents are optimized for specific tasks such as airdrops, trade sniping, and arbitrage. Griffain enhances the operability and user experience of the Solana ecosystem through these diversified features.

Currently, Griffain is in an invite-only access phase, limited to users holding the Griffain Early Access Pass or Saga Genesis Token, and adopts a SOL billing model covering transaction fees, agent service fees, etc. The platform's AI agents can provide market analysis, trading signals, automated trading strategies, and other value-added services, while users holding Griffain tokens can unlock more advanced features. As a pioneer of AI agents in the Solana ecosystem, Griffain aims to drive the wave of "Agentic App SZN" and will continue to deepen the application of AI technology in on-chain trading, market analysis, and DeFi, providing users with a smarter and more efficient crypto experience.

2. AI Influencer

AiDOL is a typical representative of the AI Influencer trend. AiDOL combines AI-generated content (AI GC), virtual image modeling, and interactive live streaming technology to create a highly influential AI idol ecosystem. Among them, Luna is the most popular AI agent, attracting a large number of fans with its highly intelligent interaction and personalized content; Iona and Olyn have also attracted many users with their unique styles and innovations. AiDOL primarily uses TikTok live streaming as its stage, accumulating 672,100 subscribers and nearly 10 million likes in a short time with high-quality short videos generated by AI and real-time interactive live streaming, becoming an important participant in the AI influence economy.

Aixbt: Automated AI Influencer

Aixbt is an AI-driven crypto market intelligence agent launched in November through Virtuals, led by developer Alex, who goes by the pseudonym @0rxbt. Alex has focused on developing analytical tools since 2017 and began exploring AI Agents applications in 2021. AI XBT is the only tokenized project owned by the developer, with 14% of tokens held by Alex locked for 6 months, which will later be used for team expansion and project development. The team has already hired UI/UX engineers to optimize terminal functions and introduced AI researchers to enhance agent intelligence. AI XBT relies on the meta-llama/Llama-3-70b-chat-hf model to achieve conversational AI, situational awareness, sentiment analysis, and retrieval-augmented generation (RAG) capabilities, ensuring efficient and accurate information processing.

AI XBT aims to create a fully automated AI influencer, using intelligent analysis tools to monitor Crypto Twitter and market trends in real-time, providing users with data-driven market insights and investment advice. Its core functions include KOL monitoring (covering over 400 key opinion leaders), blockchain data analysis, market trend forecasting, and automated technical analysis and strategic recommendations. Additionally, AI XBT shares some analysis content publicly on Twitter, while in-depth reports are accessible only to token holders. Users can also interact directly with AI through a dedicated terminal to obtain personalized investment advice and risk assessment reports. Daily, AI XBT publishes market insights at a fixed frequency and automatically replies to over 2,000 mentions to efficiently interpret market sentiment and narrative trends.

AI XBT offers two main usage methods: first, users can @AI XBT on X (Twitter) to ask questions, such as token compatibility or project metrics, and AI will analyze and respond instantly; second, the Aixbt Terminal, positioned as a "narrative analysis-driven market intelligence platform," provides deeper data analysis and strategic recommendations. Currently, this terminal is only open to users holding over 600K $AI XBT tokens, with plans to expand coverage in the future to meet market demand.

3. Dev Utility

Dev Utility refers to tools or functions that provide convenience for developers and improve productivity, especially in the fields of AI, blockchain, and Web3. It encompasses basic development tools such as code editors, debugging tools, version control, and automation tools, as well as SDKs, APIs, and smart contract development frameworks related to AI and blockchain development. In the AI & Web3 domain, Dev Utility may also involve AI agent-assisted analysis, retrieval-augmented generation (RAG), and other technologies to help developers build applications more efficiently. Its core value lies in enhancing development efficiency, optimizing workflows, and reducing development difficulty, allowing developers to focus on core business logic.

SOLENG: Code "Review"

SOLENG (@soleng_agent), as a solution engineering and developer relations agent, aims to bridge the gap between technical teams and broader project needs. Its core function is to automatically review the code submitted by participants in hackathons and provide preliminary review comments. Although robotic reviews cannot fully replace human input, SOLENG can effectively filter out obvious errors and improve review efficiency as a "juror."

The project has publicly shared review results on GitHub (link), showcasing SOLENG's role in the hackathon review process. In addition to basic pros and cons analysis, SOLENG also checks for spelling errors in the code and provides correction suggestions, making the review more practical. This model aligns with hackathon needs, providing developers with immediate feedback.

The developer behind SOLENG is Lost Girl Dev, whose identity resonates with the project's virtual female persona. Her technical capabilities have garnered attention from the official ai16z account and have interacted with Shaw on the X platform, further enhancing SOLENG's industry influence.

4. Investment DAO: Intelligent Investment Research

Investment DAO provides users with more refined investment analysis services through "investment research-type" AI agents. Its core functions include automatically interpreting candlestick charts, assisting in technical analysis, assessing whether projects have Rug risks, and generating information summaries similar to research reports. This AI-driven intelligent investment research model lowers the analytical threshold for users, enabling investors to obtain market insights more efficiently and providing strong support for decision-making.

Vader AI: AI Agent Investment DAO

Vader AI aims to become the "BlackRock" in the Agentic economy, attracting and promoting its self-trading AI Agent tokens to its followers. The platform builds a multifunctional AI Agent investment ecosystem by profiting from investments and airdropping profits to holders and followers. Its core goal is to establish itself as a leading AI Agent investment DAO management platform, driving industry innovation and scalability.

Vader AI promotes the integration of technology and capital through a multi-agent system, committed to establishing an investment DAO ecosystem managed by AI Agents. In this network, agents can not only raise funds and manage capital but also hire other agents to optimize investment strategies, enhancing the system's efficiency and flexibility. Through decentralized computing, agents can also reinvest in research and development, promoting the platform's continuous growth.

Additionally, Vader AI adopts an innovative token incentive mechanism to provide B2B tool optimization for investors, enhancing the platform's commercial application value. The platform further solidifies investors' sense of participation and profit-sharing mechanisms by sharing GP/carry profits with holders, making Vader AI not only an investment platform but also a multi-win ecosystem empowering agents and investors.

5. Content & Creator

Whether in writing, editing, or visual design, AI can provide personalized creative outputs based on user needs, helping creators save time, enhance productivity, and stand out in fierce market competition. The platform aims to provide content creators with an intelligent and convenient creative assistant, promoting innovation and development in the content industry.

ZEREBRO: AI Art Creation and Content Generation

ZEREBRO is a blockchain-based cross-chain natural intelligence autonomous AI agent focused on art creation and content generation. Its innovative combination of decentralized verification, meme generation, NFT minting, and DeFi applications demonstrates strong multifunctionality and execution capability. ZEREBRO has successfully operated Ethereum mainnet verification nodes and sold artworks on Polygon, accumulating important assets for its economic foundation.

ZEREBRO is also committed to building a decentralized computing network and implementing MEV optimization strategies to ensure economic and technical sustainability. It is not only a technical tool but also explores the deep involvement of agent technology in blockchain operations, economic models, and governance. ZEREBRO promotes its value in decentralized ecosystems through multiple dimensions.

ZEREBRO tokens have two main uses: first, as content interaction rewards, allowing token holders to earn by participating in decentralized content on social platforms; second, as community development tools, rewarding users who actively participate in the ecosystem, including content creation, staking, and governance, further enhancing community activity and participation.

6. Gaming & Agentic Metaverse

Gaming & Agentic Metaverse is exploring AI-driven gaming and metaverse experiences, dedicated to creating a virtual world where humans interact with agents through reinforcement learning. This emerging field combines artificial intelligence with immersive gaming environments, allowing players to dynamically interact with intelligent agents and experience more personalized and intelligent gameplay.

ARC: AI Solution Provider

ARC addresses player liquidity issues in independent games and Web3 games through AI technology. The project has upgraded from a single game studio (AI Arena) to a comprehensive AI solution provider, launching ARC B2B and ARC Reinforcement Learning (ARC RL). ARC B2B is an AI-driven game development toolkit (SDK) that can be seamlessly integrated into various games, providing developers with intelligent gaming experiences. ARC RL trains "super intelligent" game agents using crowdsourced game data through reinforcement learning to enhance gameplay and sustainability. ARC's business model is deeply tied to integrated game studios, with revenue sources including token distribution in Web3 games and royalties based on game performance, while building a generalized AI data reserve across game types to promote the training and evolution of general AI models.

ARC's technical applications cover multiple core modules. AI Arena is a cartoon-style AI competitive game where players train AI warriors for combat, with each character being an NFT, enhancing the game's strategic and economic value. ARC SDK allows developers to easily integrate AI agents, deploying models with just one line of code, while ARC handles backend data processing, training, and deployment. ARC RL enhances AI training efficiency through offline reinforcement learning, allowing agents to learn from human player data, thus providing more natural and challenging game opponents. ARC's AI model architecture includes feedforward neural networks, table agents, hierarchical neural networks, etc., to adapt to the interaction needs of different types of games while optimizing state and action spaces to ensure smooth and intelligent gaming experiences.

ARC covers both independent games and Web3 games, helping developers solve early player liquidity issues and enhancing the long-term attractiveness of games. The core team members have rich experience in machine learning and investment management, securing $5 million in seed funding led by Paradigm in 2021 and an additional $6 million follow-up funding in 2024. ARC's native token NRN has transitioned from a single game economy (AI Arena) to a platform economy, adding demand-driven factors such as integrated revenue, Trainer Marketplace fees, and ARC RL staking participation, ensuring the sustainability and value growth of the token. Through a crowdsourced data contribution mechanism, ARC RL achieves collaborative training among multiple participants, promoting the intelligent evolution of AI agents and further enhancing the vitality and competitiveness of the gaming ecosystem.

7. Framework & Hubs

When developing AI Agents in the crypto space, many frameworks, while suitable for basic projects or toy-level applications, often expose issues of insufficient customization and excessive abstraction complexity in real product development, making it difficult for developers to flexibly expand and apply. Excellent Agent frameworks need to address core pain points, including: comprehensive support for on-chain operations, efficient integration of on-chain data, DeFi automation, NFT, and other key application scenarios' APIs; multi-platform compatibility, supporting major blockchains and social platforms to achieve integrated user operations; modularity and flexibility, abstracting basic functions, such as vector storage and LLM model switching, allowing developers to flexibly adapt to different needs and avoid redundant development; memory and communication capabilities, although some frameworks invest significant resources to enhance this capability, excessive intelligence at the current stage may not be practical and may instead increase complexity.

The following is a detailed comparison of mainstream crypto AI Agent frameworks in various dimensions:

Eliza ($AI16Z): AI Agent Framework

Eliza ($AI16Z) occupies a leading position in the AI agent market, with about 60% market share and a strong TypeScript ecosystem, attracting numerous developers. Its GitHub project has accumulated over 6,000 stars and 1.8K forks, demonstrating high community engagement. Eliza excels in multi-agent systems and cross-platform integration, supporting mainstream social platforms such as Discord, X (Twitter), and Telegram, making it an important player in the social AI and community AI fields. With a broad ecological foundation, Eliza has excellent adaptability in social interaction, marketing, and AI agent development.

In terms of technical architecture, Eliza has multi-agent system capabilities, allowing different AI roles to share runtime environments and achieve more complex interaction patterns. Its retrieval-augmented generation (RAG) technology endows AI with long-term contextual memory capabilities, enabling it to maintain consistency in continuous conversations. Additionally, the plugin system supports extensions such as voice, text, and multimedia parsing, further enhancing the flexibility of application scenarios. Eliza is also compatible with multiple LLM providers such as OpenAI and Anthropic, providing efficient AI computing capabilities whether deployed in the cloud or locally. With the launch of the V2 message bus, Eliza's scalability will be further optimized, making it suitable for medium to large social AI applications.

Despite Eliza's outstanding performance in the market, it still faces certain challenges. Its multi-agent architecture may lead to complexity issues in high-concurrency scenarios, increasing system resource overhead. Furthermore, the current version is still in early development, with stability and optimization continuously improving. For developers, the learning curve of the multi-agent system is relatively steep, requiring a certain level of technical accumulation to fully leverage its advantages. In the future, with continued community contributions and the release of the V2 version, Eliza is expected to achieve further breakthroughs in scalability and stability.

GAME ($VIRTUAL): AI Agent Framework

GAME ($VIRTUAL) focuses on gaming and the metaverse, significantly lowering the development threshold for developers through low-code/no-code integration, enabling them to quickly build and deploy intelligent agents. Additionally, leveraging the $VIRTUAL ecosystem, GAME has formed a strong developer community, accelerating product iteration and ecological expansion. Its core advantage lies in providing efficient game AI solutions, making procedural content generation, dynamic adjustment of NPC behavior, and on-chain governance easier to implement.

In terms of technical architecture, GAME adopts an API + SDK model, providing convenient integration methods for game studios and metaverse developers. Its agent prompt interface optimizes the interaction between user input and AI agents, making intelligent behavior in games more natural. The strategic planning engine divides the logic of AI agents into high-level goal planning and low-level strategy execution, enabling stronger adaptability in complex gaming environments. Furthermore, GAME supports blockchain integration, allowing for decentralized agent governance and on-chain wallet operations, giving it a unique advantage in the Web3 gaming field.

GAME has optimized performance for high-concurrency gaming scenarios, performing well in handling game engine constraints. However, its overall performance is still affected by the complexity of agent logic and the overhead of blockchain transactions, which may pose challenges to real-time interactivity. Additionally, as an AI agent framework focused on gaming and the metaverse, GAME has limited versatility in other fields. The complexity of blockchain integration still needs optimization to reduce development costs and further attract a broader developer community.

Rig ($ARC): AI Agent Framework

Rig ($ARC) holds a 15% market share in the enterprise-level AI agent market, with high performance and modular architecture based on the Rust language, excelling in high-throughput and low-latency scenarios, especially suitable for high-performance blockchain ecosystems like Solana. With strong system stability and efficient resource management, Rig is an ideal choice for on-chain financial applications, large-scale data analysis, and distributed computing tasks. Its architectural design emphasizes scalability, allowing enterprise users to flexibly deploy AI agents in complex data environments, improving computational efficiency.

In terms of technical architecture, Rig adopts a Rust workspace structure, ensuring code modularity and readability while enhancing system scalability. Its provider abstraction layer supports seamless integration with multiple mainstream LLM providers (such as OpenAI and Anthropic), allowing developers to switch models freely. Rig also supports vector storage, compatible with backend databases like MongoDB and Neo4j, improving contextual retrieval efficiency. Additionally, Rig has a built-in agent system that combines RAG models and tool optimization capabilities, enabling it to automate complex task execution, suitable for high-performance computing and intelligent data processing scenarios.

Rig achieves excellent concurrency performance through Rust's asynchronous runtime, capable of scaling to high-throughput enterprise-level workloads. However, the steep learning curve of Rust itself may pose an entry barrier for some developers. Furthermore, Rig's developer community is relatively small, and the ecological driving force still needs to be strengthened. Nevertheless, with the growing demand for Web3 and high-performance computing, Rig still has broad market potential and is expected to further enhance market penetration by optimizing developer experience and strengthening community building.

ZerePy ($ZEREBRO): AI Agent Framework

ZerePy ($ZEREBRO) occupies a 5% market share in the creative content and social media automation field, with a total market value of $300 million. Its core advantage lies in a community-driven innovation ecosystem, accumulating a loyal user base in applications such as NFTs, digital art, and social content automation. ZerePy lowers the development threshold for AI agents, enabling content creators and community operators to easily deploy intelligent agents for automated content creation, social interaction, and community management, enhancing user engagement and content influence.

In terms of technical architecture, ZerePy is based on the Python ecosystem, providing a friendly development environment for AI/ML developers, while leveraging the modular Zerebro backend to achieve agent autonomy in social tasks. Its social platform integration optimizes Twitter-like interactions, allowing agents to automatically perform tasks such as posting, replying, and retweeting, enhancing social media automation capabilities. Additionally, ZerePy's lightweight architectural design makes it more suitable for individual creators and small community AI agent needs without incurring high computational costs.

ZerePy performs well in social interaction and creative content generation, but its scalability is primarily suitable for small-scale communities and less so for high-intensity enterprise-level tasks. Moreover, due to its concentrated application scope, its applicability outside the creative field still needs further validation. For scenarios requiring more complex creative outputs, ZerePy may require additional parameter tuning and model optimization to meet broader market demands. With the development of the creative economy, ZerePy is expected to further expand application scenarios in NFT generation and personalized social agents.

8. AI Launchpad

AI Launchpad not only provides emerging projects with customized growth paths, covering technical support, fundraising, marketing, and collaboration opportunities with industry experts, but also helps projects quickly integrate into the global AI community through its extensive collaboration network.

Vvaifu: The First AI Launchpad on the Solana Chain

vvaifu.fun is the first AI agent launchpad based on the Solana chain, allowing users to create, manage, and trade AI agents without any coding skills. The platform enables each AI agent to have its own token, forming a decentralized ecosystem. Users can not only co-own these agents but also interact with AI-driven assets. The platform supports agents' autonomous interactions on social media platforms such as Twitter, Discord, and Telegram, and features on-chain wallet management, greatly enhancing its practicality in various application scenarios.

The business model of vvaifu.fun is based on its unique token economic model. The platform's main token, $VVAI FU, is the first AI agent token launched on the Dasha platform, featuring deflationary characteristics, with a certain amount of $VVAI FU burned each time an agent is created or a function is unlocked. Additionally, the platform has designed multiple burning mechanisms to ensure token value stability, including burning 750 $VVAI FU when an agent is created, consuming $VVAI FU and SOL fees when functions are unlocked, etc. Each launched agent will also allocate 0.90% of the new agent tokens to the community fund or directly into the team treasury, promoting community participation and ecosystem building.

The platform's community participation mechanism enhances user interactivity and governance rights. Token holders can accumulate 0.90% of the supply initiated by agents through the community wallet and vote on the use of these resources. vvaifu.fun has also set the platform transaction fee at 0.009 SOL, providing sustainable economic support for the platform's operations. Through these mechanisms, vvaifu.fun provides a comprehensive decentralized interactive platform for AI agent creators and users, not only promoting the development of creative projects but also incentivizing active participation from the global community.

Clanker: AI Reply Bot

Clanker is an AI reply bot based on Farcaster, designed for users to create and deploy memecoins and tokens. Through this platform, users can easily create their own tokens by interacting with Clanker. Users simply need to tag @clanker on Farcaster, informing the bot what kind of token they need and providing details such as name, code, image, and supply. Clanker will generate and provide a tracking link within a minute, ultimately deploying the token on Uniswap v3, although without initial liquidity, users need to manually add liquidity to price the token.

The underlying technical architecture of Clanker operates by combining Next.js middleware with LLMs (such as Anthropic's Claude or ChatGPT). When users initiate a request on Farcaster, the message is forwarded to the LLM, which executes decision logic based on the provided context to determine the token deployment operation. This process illustrates how Clanker leverages AI technology to simplify the user-generated and deployed token process, fully integrating social platform and blockchain technology to provide users with a convenient token creation experience.

As a platform, Clanker not only simplifies the creation process but also deeply integrates with Uniswap v3, allowing users to directly deploy new tokens to decentralized exchanges. This process accelerates the issuance of memecoins and tokens and supports strategic value for the ecosystem through components such as Telegram bots, DEXs, and aggregators, thereby driving the growth of on-chain trading volume. With the increase in the number of tokens, Clanker has participated in a significant rise in trading volume, helping users leverage the advantages of low transaction fees and fast confirmation times, promoting the circulation of on-chain assets like Solana and Base.

Key Conclusions

Technology-driven and infrastructure forms the core of AI agent projects, ensuring efficient operation and supporting scalable expansion through advanced programming languages and innovative algorithms. At the same time, high-performance blockchain platforms provide excellent transaction processing capabilities and multi-chain compatibility, enabling AI agents to interact seamlessly across different chains, driving continuous optimization and upgrading of the technological foundation.

Payment and trading infrastructure are key pillars of AI agent ecosystem development. Stablecoin payment systems ensure transaction stability and liquidity, enhancing the interaction efficiency between AI agents and users. Decentralized autonomous trading systems achieve more efficient and secure automated trading by eliminating human intermediaries. Additionally, innovative reward and governance mechanisms, such as "proof of contribution" and "proof of collaboration," promote AI agent collaboration and resource sharing, ensuring the long-term healthy development of the ecosystem through a sound governance system.

Outlook and Challenges

The necessity of AI Agent tokens is often questioned, mainly because they do not directly enhance the functionality of agents or bring obvious advantages. Many believe that AI Agent tokens are similar to tokens in Web3 games, which may not substantively help the core functions of the projects. Therefore, some investors may overlook the actual value of these tokens due to blindly following the AI trend, leading to high risks and even potential scams. For such projects, some believe they attract unsuspecting investors by disguising legitimacy, especially compared to meme coins, which may promise too many unrealized functions behind these tokens.

If a project prioritizes tokens as the primary driving force, it may sacrifice core functions and experiences, especially in non-gambling nature games and services. Tokens should serve as supplementary elements rather than dominant factors. Many successful projects have proven that truly effective applications should focus on user experience, creating high-quality products rather than merely relying on token economic incentive mechanisms to attract users.

The integration of AI and DeFi will be an important trend in the future, with an expected 80% of DeFi transactions to be completed by AI Agents, with promoters like Modenetwork and Gizatech actively driving this development. At the same time, the role of AI Agents in protocol governance will be further expanded, potentially even triggering AI-driven governance attacks. Additionally, security-oriented AI Agents are expected to play a crucial role in protecting protocols from attacks, similar to the protective functions provided by HypernativeLabs and FortaNetwork. With the continuous expansion of infrastructure, the development of Trusted Execution Environments (TEE) and the core position of decentralized computing will enhance the resilience of AI Agents. Furthermore, the explosion of AI data markets will also drive the growth of data payments between AIs, with projects like Nevermined.io laying the groundwork for this.

Disclaimer

This content is for reference only and does not constitute or should not be regarded as (i) investment advice or recommendation, (ii) an offer or solicitation to buy, sell, or hold digital assets, or (iii) financial, accounting, legal, or tax advice. We do not guarantee the accuracy, completeness, or usefulness of such information. Digital assets (including stablecoins and NFTs) are subject to market fluctuations and involve high risks, which may depreciate or even become worthless. You should carefully consider whether trading or holding digital assets is suitable for you based on your financial situation and risk tolerance. Please consult your legal/tax/investment professionals regarding your specific circumstances. Not all products are available in all regions. For more details, please refer to OKX's Terms of Service and Risk Disclosure & Disclaimer. The OKX Web3 mobile wallet and its derivative services are governed by separate terms of service. You are responsible for understanding and complying with applicable local laws and regulations.