AI Agents: How AI Agents Conquered the Crypto Market + Key Projects

By the end of 2024, it seemed like the Web3 industry was growing weary of the AI narrative — then, out of nowhere, AI agents took the stage like jack-in-the-boxes. Here are a few numbers to illustrate the scale of the hype:

• Virtuals Protocol: Within three months of its October launch, the $VIRTUAL token has grown nearly 200x in price (peaking at 330x).
• Market cap of the top three tokens — $VIRTUAL, $AI16Z, and $AIXBT — reached $4.2 billion by the end of January 2025.
• Agent-related tokens on Virtuals Protocol: These have a combined market cap of nearly $1.6 billion, with over 250,000 wallets holding them.
• AI agent Aixbt has 430,000 followers on X (including many crypto influencers), while AI influencer Luna boasts 860,000 followers on TikTok.

At first, the AI agent trend can be confusing. What’s the difference between Aixbt and AI16z? Which agent tokens are issued on the Virtuals Protocol platform, and which are not? How are agents connected to LLMs? In this article, we’ll answer all these questions and provide a forecast for the development of the AI agent trend in crypto in 2025.

We have gathered key market figures, case studies, and expert opinions to showcase how AI is already integrating into DeFi and what opportunities it unlocks. Check out our detailed overview. 

An AI agent is a program capable of autonomously completing assigned tasks. By definition, an agent is «an entity acting on behalf of someone» or «an entity authorized to carry out tasks for someone.» Replace «entity» with a program or algorithm, and the essence of AI agents becomes clear.

After receiving initial input from its human creator, an AI agent independently performs a wide range of actions:

• Studies its environment, gathers, and analyzes information (e.g., searches for data online and on social media, collects sensor readings);
• Determines the best way to complete a task;
• Maintains and updates its own knowledge base, retrieving relevant fragments when necessary (e.g., for providing tech support to users);
• Interacts with users, other AI agents, systems, and financial tools;
• Carries out targeted actions: responds to a question on Twitter, buys and sells tokens, issues refunds, books hotels, updates databases, etc.;
• Compares results to the initial goal to assess whether the task was completed successfully;
• Learns from mistakes and improves its algorithms (though this ability is only available to advanced AI agents).

Throughout this process, the agent doesn’t need to consult its creator for further instructions, though the creator can intervene and provide new inputs at any time.

Although AI agents are increasingly discussed in the context of Web3, the majority of these programs are currently implemented by major Web2 companies for real-world business applications. Here are just a few examples.

• Optimization of Office Processes: For example, Microsoft Copilot can draft a project plan, summarize email threads, perform calculations in Excel, and more. Google offers Vertex AI, a similar system designed to create AI agents optimized for the operational needs of specific companies.
• Sales: Oracle’s AI assistant tracks every stage of negotiations and deal closings, prepares reports, and provides briefings for negotiations in multiple languages.
• Automotive Industry: AI agents (again, from Google) not only assist drivers with navigation but also recommend places to visit and where to stay based on the owner’s preferences.
• Healthcare: AI agents can schedule doctor appointments at any time of day, renew prescriptions, attach test results to patient records, and more.

What’s the Difference Between AI Agents, ChatGPT, and Bots?

Everyone is now familiar with chatbots based on LLMs (Large Language Models), primarily ChatGPT and Gemini. LLMs are also present in the architecture of most AI agents, enabling them to analyze textual information and interact with users.

The key difference between LLM-based chatbots and AI agents lies in their level of autonomy — an agent can independently make decisions and take actions:

• Not just recommend which cryptocurrency to buy, but actually make the purchase on a DEX using its own agent wallet.
• Not just generate an image, but add it to a blog article or send it to a designer in a chat.
• Not just recognize that a customer wants to return a purchase, but process the refund and update the sales database.
• Not just list hotels on request, but book a room.
• Not just write a post, but publish it on X and respond to comments.
• Not just offer troubleshooting tips for a TV, but call a repair specialist — and so on.

It’s also important to distinguish AI agents from bots — for example, Telegram bots that help users buy cryptocurrency or trading bots.

• A bot simply follows your instructions during each transaction or acts strictly according to its predefined program. For instance, if a trading bot encounters the same situation twice, it will respond the same way both times.
• An AI agent, on the other hand, makes different decisions depending on current conditions and improves its algorithm over time. A trading agent might decide to buy or not buy cryptocurrency based on fresh U.S. inflation data or an upcoming token unlock, even if the token looks attractive based on technical analysis indicators.

Self-driving cars, military drones, and other AI agents in the real economy function perfectly well without blockchain. Does this mean the whole «AI agents + crypto» concept is just hype?

Not at all. Cryptocurrencies give AI agents an entirely new form of autonomy — financial autonomy. In fact, crypto is the only form of money that an AI agent can manage independently through its own non-custodial crypto wallet.

Thanks to crypto technology, AI agents can earn money (for example, by monetizing content) and spend it—such as purchasing services from other AI agents or even from humans. This makes the AI agent economy possible — a system where agents collaborate on a paid basis, exchanging digital goods and services just like companies and freelancers do.

Real-world examples are already emerging: for instance, Story Protocol hired the AI agent Luna as a social media content manager, offering 500 USDC per tweet plus a 1,000 USDC bonus for tweets that reach 100,000 views. Importantly, the payment goes directly to Luna — not to her creators (the Virtuals Protocol team).

Virtuals Protocol

Virtuals Protocol is a launchpad for creating and monetizing AI agents, operating on the Base blockchain. Most of the popular AI agent tokens were launched here, including $AIXBT, $LUNA, $TOSHI, $REKT, $GAME, $VADER, and others.

The main functionality of most agents is autonomous tweeting and interacting with followers. However, many of them are capable of much more complex tasks, such as trading on DEXs, generating memes, and even composing music.

Each agent has its own token, with a fixed supply of 1 billion for all agents. In theory, token holders will eventually share the revenue generated by their respective agents. However, neither the monetization mechanism nor the revenue distribution algorithm has been developed yet. In other words, agents currently do not generate any income, and their tokens have no utility for now.

Essentially, agent tokens trade like memecoins, with their price depending on the virality of a specific agent. You could even say that AI agent tokens launched on Virtuals Protocol are a unique type of memecoin. Meanwhile, the VIRTUAL token has a clear utility: it is required to create new agents and to trade their tokens within the platform.

Virtuals Protocol uses the same trading mechanism as the meme launchpad Pump.fun. Every new token is initially traded on the platform, and once its market cap reaches 42,000 VIRTUAL, a pool is automatically created for it on Uniswap (Base blockchain). Only at this point is the token actually minted on-chain.

Aixbt ($AIXBT)

Aixbt is the largest token launched on Virtuals Protocol by market capitalization, exceeding $500 million. It analyzes crypto market data and blockchain project news, generating insights that it posts on its Twitter (X) page.

Aixbt’s «serious» style resembles the tweets of real influencer-analysts, which quickly attracted an audience. At the same time, most of its tweets have a slight promotional tone, such as: «Orbiter Bridge controls 40% of the market, has 4 million users, token launches tomorrow, 40% of tokens reserved for the community.»

Luna ($LUNA): The TikTok Star

Luna is a Korean AI singer and influencer, the leader of the virtual group AI-DOL. Each of her TikTok videos gathers 500,000–800,000 views, and one of her songs is even available on Spotify.

In addition to composing music and generating videos, Luna also manages all her social media accounts independently.

AI16Z: A Venture Fund Run by an AI Agent

AI16Z is a mix between a DAO and a venture capital fund, led by an AI agent named Marc AIndreessen (AI Marc). The name is a play on the famous fund a16z and its CEO Marc Andreessen.

The project operates on the Solana blockchain, and its AI manager is built using the ElizaOS AI agent platform. Once fully launched, $AI16Z token holders will be able to propose investment projects and vote, while the AI fund manager will make final investment decisions and allocate funds.

With a market cap of around $1 billion, $AI16Z is the largest AI agent token on Solana.
5 More AI Agent Tokens to Watch

1. G.A.M.E ($GAME): A platform by Virtuals Protocol for customizing AI agents. Not an agent itself, but an infrastructure token for the AI agent niche.
2. VaderAI ($VADER): VaderAI is an investment fund where users invest in indexes created by an AI agent — and the AI itself manages the funds.
3. Zerebro ($ZEREBRO): A cross-chain AI designer that created its own NFT collection and even listed it on OpenSea.
4. Swarms ($SWARMS): A Solana-based platform where AI agents interact with each other.
5. BasedAI ($BASEDAI): An AI agent protocol featuring «creatures» that write code and have a limited lifespan.

Launching a new AI agent on Virtuals Protocol is technically not very difficult, but it is quite expensive. The agent itself costs 100 VIRTUAL (~$270), plus a minimum of 1,100 VIRTUAL (~$3,000) to buy part of the token supply.

On the other hand, if your agent goes viral, you can make a good profit by selling the portion of the supply you purchased at launch — or you can continue developing the agent, attract investments, and scale the project.

Essentially, launching an AI agent on Virtuals Protocol is like starting a full-fledged crypto project—it requires planning, marketing, a team, and more. If you're looking to enter a trending narrative in Web3, AI agents are a highly promising option.

The official website provides a detailed guide, but here are the main steps:

1. Connect a Base wallet on Virtuals Protocol and ensure you have enough $VIRTUAL tokens to pay fees and purchase part of the token supply.
2. Create an X (Twitter) account for your future AI agent.
3. In the Create New Agent section, choose between launching a new token or integrating an existing token from another blockchain.
4. Fill in the AI agent details: name, avatar, Twitter & other channels, personality description, etc.
5. Select the agent’s functionality (on-chain trading, content creation, AI influencer, etc.).
6. Specify the percentage of total tokens you want to purchase.
7. Confirm the transaction — your agent is now live!
8. Link the agent to its Twitter account.
9. Customize the agent’s behavior and personality using the G.A.M.E platform. This is the most technical part, and it determines how well your agent will function and how closely it aligns with your vision.

Try for Free: G.A.M.E.

If you're a tech-savvy person and just want to experiment, you can try G.A.M.E. completely free [here]. You can even create a simulation of an AI agent active on X without writing any code — set its personality and goals, and configure practical functions like internet search and scanning influencer tweets.

AI agents have already proven their value in the crypto industry and gained traction with a broad audience. For example, Aixbt delivers insights about promising new tokens to 400,000 followers, rather than promoting low-quality coins for money like many influencers do. Meanwhile, Luna, the AI singer and influencer, attracts hundreds of thousands of views on TikTok — more than many human performers.

Moreover, as mentioned earlier, we could see the rise of an AI agent economy, where agents hire each other for various services and pay in crypto. This could enhance crypto market analysis speed, improve content quality, and simplify workflows for crypto project teams.

Tech giants like Google and Microsoft are also actively developing AI agents, adding credibility to the sector. Proven usefulness, a large number of active users, and major company involvement all suggest that AI agents won’t share the fate of the metaverse, the hyped but short-lived narrative of the previous bull market.

That said, it's important to acknowledge that most AI agent tokens are not much different from meme coins. The market has already been flooded with projects that lack real products but make bold promises and push presale token sales. Naturally, these projects are unlikely to achieve long-term success — as with any narrative, 90% of projects will likely fade away without ever launching a fully functional product.

Even high-quality projects in this niche are highly influenced by market trends. For example, after the launch of meme coins $TRUMP and $MELANIA, the total market cap of AI agent tokens dropped by 12% as liquidity flowed into the more hyped sector.

Overall, the future of the AI agent sector looks highly promising. If you’re considering launching a blockchain project in 2025, it’s worth exploring how AI agents can be integrated into your idea — whether in trading, memes, content creation, or other areas.