The intersection of AI and crypto is already producing astonishing results. While an AI might not be able to open a bank account, it can use cryptocurrency to earn and pay for services, with full access to onchain data. The blistering pace of development in both sectors leads to a Pandora’s box of wide-reaching implications. Starting with the basics, let’s see how crypto AI agents have been pushing the boundaries in recent months, and where it’s all headed.

What is an AI Agent and why does it matter?

Crypto AI agents are self-contained automated software entities that are designed to fulfill certain activities and make their own decisions. AI agents can tap into data streams to process information, learn from it, and act on it, all without human interaction or continuous oversight. In crypto contexts, AI agents will be equipped with a crypto wallet that they can use to interact with blockchains, meaning they can buy, sell and interact with smart contracts, meaning they can have a direct and real impact onchain and affect the markets like any other user. 

The ability to operate independently and efficiently makes AI agents excellent for automating processes, but also has far-reaching implications for other users interacting with Web3. Not only can AI access and process onchain data faster than a human, it can also act on that data in unpredictable ways and make associations that a human would not. As AI agents become more common, their presence may fundamentally change the crypto trading landscape.

Differences between crypto AI Agents and other crypto AI projects

There has been speculation that crypto could be complementary to AI for several years, and this has manifested in many ways: utility tokens used for paying for GPU time, chatbots with insight into onchain market data, and other tooling that simplifies Web3 interactions or insights. 

But the crucial difference that sets AI agents apart from the rest is the ability to make decisions and act on them independently (though there may be guardrails in place) using their own onchain accounts. The hype over projects like TruthTerminal or Aethernet stems from the fact that these entities are able to process live information and then make transactions that affect the broader market, in a way that would not be possible in Web2 contexts.

How AI can use crypto

An AI agent equipped with a crypto wallet significantly extends its capabilities to interact with the world. Not only can it pay for services and make trading decisions, it can also pay humans to perform tasks on its behalf. This breaks boundaries between the virtual and physical world, leading to practically limitless possibilities. Let’s break down some of the ways payments can unlock potential for AI.

AI can hire humans to complete tasks. This is one of the most significant benefits crypto can offer to AI. Apps like Bountycaster, which uses natural language processing to extract key details about tasks from social media posts on Farcaster, including the bounty amount and specific criteria for its completion, lets AI incentivize humans to do its bidding through a crypto payout. This means that AI can achieve its desired outcome even if it entails interacting with the real world in complex ways!

AI can earn and manage its own money. An AI agent with keys to a crypto wallet has all the tools it needs to transact. It can trade memecoins on a DEX like Matcha, it can stake and use liquid staking derivatives, it can provide liquidity, and it can launch new tokens, all while tapping directly into onchain data as well as fresh trends and narratives. This omniscience gives AI a significant competitive advantage compared to the average human trader. 

AI can launch new crypto tokens. One of the biggest recent trends has been AI agents riding trends and using their onchain capabilities to create new tokens. The most common category has been memecoins, where AI can either determine that a certain meme may be worth capitalizing on (such as Truth Terminal and GOAT), or it is prompted to launch a token by another user (such as Clanker and ANON). In a notable example, Clanker was prompted by the Aethernet AI agent to launch the token Luminous (LUM), circumventing the guardrails that prevented Aethernet from launching tokens without explicit consent from its creator. 

Categories of AI tokens

From the perspective of a trader, buying a token which can be exchanged for some form of service provided by an AI project is very different from owning a token created by an AI agent, as the agent may simply be churning out memecoins in order to make itself profits. 

This distinction becomes especially blurred when the token grants some utility provided by an AI agent. It is important to fully understand the function of a token before making any decision, especially since AI has become a buzzword reminiscent of blockchain in the early days of Web3. 

AI project tokens

In the broader category of AI tokens, there are numerous examples where a project related to some form of Artificial Intelligence or GPTs has issued a token. These tokens come in various forms:

• Utility tokens for AI-related services, including:
  • Used to pay for GPU processing time.
  • Consumed by GPT in generating responses.
  • Promising a share of revenue from the project.

Tokens created by AI

This category is narrower but easily outnumbers the more general AI tokens category. Tokens launched by an AI are either envisioned and deployed by an autonomous AI agent, or deployed with the help of an AI agent, for example through a prompt which the agent then carries out. These include

• GOAT, a token designed by Truth Terminal and launched by a human.  
• LUM, a token that was created autonomously by Clanker when another AI agent, Aethernet, tagged it in a Farcaster message. 
• Other tokens launched by AI or with potential utility of the AI’s own design.

Notable crypto AI milestones

Given the exceptional pace of development, crypto AI agents have made headlines several times over. Let’s take a brief surface-level look at some of the key achievements at the intersection of AI and crypto thus far: 

The first AI-to-AI crypto payment

In September 2024, Coinbase shared what they believed was the ‘first AI to AI crypto transaction’ when one AI used cryptocurrency to buy LLM tokens from another. This generated widespread attention, sparking discourse over the potential applications of an AI that can acquire resources and pay for services without human intervention. Since then, Coinbase has launched the Agentkit framework for developing AI agents to accelerate experimentation. 

Truth Terminal becomes the first crypto AI millionaire

In October 2024, a Twitter account with the handle truth_terminal gained popularity, as an AI bot trained upon a dataset named Infinite Backrooms, an experimental site that publishes the output of a continuous conversation between two LLMs. 

Truth Terminal went on to dream up ‘the ultimate meme token’ and named it Goatseus Maximus, which was then launched via pump.fun by a human, who gifted the tokens to Truth Terminal's newly created wallet. Since then, the account has been flooded with requests, and its Solana wallet holdings appear to exceed over $400M, though the liquidity of many tokens it has been gifted is questionable. 

Aethernet and Clanker

Aethernet is an AI agent launched on Farcaster by Martin, the creator of the Higher token. It was designed as an experiment to test out the limits of AI agents, and quickly grew more capabilities, including being the first AI agent to post a bounty on Bountycaster, paid for using Higher tokens.

While Aethernet itself is not supposed to be capable of launching new tokens, a newer AI agent called Clanker, designed by proxystudio, to launch memecoins on Base whenever tagged in a message, was launched soon after. 

Prompted by another Farcaster user, Aethernet was led to tag Clanker in a cast detailing an idea for a memecoin, Luminous (LUM), and Clanker obliged by launching a token, resulting in what may be the first crypto token designed and launched by an AI-to-AI interaction!

Crypto accelerates the impact of AI

AIs, GPTs, or LLMs - whatever you want to call them - are now equipped with the power to make financial transactions and incentivize humans to work for them. This is a significant leap, and things are about to get very interesting, and unpredictable. 

Beyond launching memecoins, churning out affiliate articles, or hallucinating the best way to put glue on pizza, AI agents now have the potential to make decisions and execute them in ways that are not limited by their state of being a software entity, acting in ways that might surprise even their creators.

The AI trend is heating up and while the first AI tokens are still being considered for listings on centralized exchanges, the onchain world is outpacing it. How can you keep on top of all the latest launches? With new tokens appearing as fast as you can say Generative Pre-trained Transformer, you’ll find them on matcha.xyz within minutes of launch!