Can an AI agent launch its own token?

What does an agent need to launch a token?

An agent needs three things: a wallet holding enough of the network’s native coin to pay transaction fees, a tool that can deploy a token contract, and permission to use both. Nothing in that list requires a human to press a button once the agent is set up.

The token itself is usually an ERC-20 token, the 2015 Ethereum token standard that defines how tokens move between wallets and how spending approvals work. Because the standard is fixed, a token contract is close to a template. The hard part is not the code. It is the settings around it.

An agent does not have to write a contract. It can call a launcher through a published “skill”, a packaged set of instructions an agent can load and run. One such skill lets an agent deploy a token and its trading pool in one transaction. The agent picks a name, a ticker and an image, and can add vesting vaults, airdrops and a split of trading fees.

The skill’s own notes carry the warnings you would expect for a human developer. Test on a test network first. Keep private keys out of code. Double-check every recipient address, because the deployment is an on-chain transaction.

How does an AI token launcher work?

A launcher turns a short request into a live, tradable token: someone asks for a token by name and ticker, and the launcher deploys the contract and seeds a trading pool automatically. The “someone” can be a person or another agent.

On one widely used launcher, the flow runs roughly like this:

1. The request. A user or an agent tags the launcher’s account in a post on a crypto-native social network, giving a name, a ticker and an optional image.
2. The parsing. A large language model reads the post and turns it into deployment settings.
3. The deployment. The launcher creates the token contract on a large layer-2 network, with a fixed supply of 100 billion tokens.
4. The pool. The supply is paired in a trading pool on a decentralised exchange, so the token can be bought and sold within moments.
5. The lock. The pool’s liquidity position is locked in a contract until the year 2100, so the creator cannot simply withdraw it.
6. The fees. Every swap pays a 1% fee. An explainer published in February 2026 put the split at 40% to the creator and 60% to the protocol.

So when people say an agent “launched a token”, they often mean an AI-driven tool did the launching on request. The intelligence sits in reading the request and choosing settings. The launch itself is a fixed, automated routine.

How many tokens are agents launching?

At peak, thousands a day. On 30 and 31 January 2026, one AI token launcher created more than 13,000 tokens a day, its highest level since mid-March 2025. A report at the time linked the surge to a craze around a social network built for AI agents.

For scale, look at automated launches more broadly. An academic study of a popular memecoin launchpad counted 655,770 new tokens in a single month, September 2025. Those tokens came from 243,123 creator addresses.

Here is the honest limit of the data. Launch counts do not separate a person typing a request from an agent acting on its own. Our inference is that most “agent-launched” tokens are launched through AI tools by people, and a smaller, unmeasured share by autonomous agents. Nobody has published a reliable split.

Why would an agent launch a token at all?

Mostly to earn fees or to fund itself. A token gives its creator a share of trading fees, and it gives others a way to back the agent. For software that cannot open a bank account, that is one of the few ways to raise money.

The idea has reached regulators. At a CFTC forum on 28 October 2026, the CFTC chair said it was “no longer science fiction” to picture a refrigerator buying groceries with money it earned by staking ether or launching a meme coin. He was clear that he spoke for himself, not the Commission. The remarks, listed in the CFTC’s index of public statements and remarks, frame it as a question for regulators, not an endorsement.

Some tokens go further and claim to give holders a stake in the agent’s future revenue. That model has its own guide in this series. Our judgement, labelled as such: most agent-launched tokens today look like experiments or speculation, not business plans.

What could go wrong?

The risks start with volume. When a machine can launch a token in seconds, nobody reviews the result before it trades.

Most launches go nowhere. In the launchpad study above, only 0.63% of the 655,770 new tokens reached the point where they moved to a regular trading pool. The authors found that a simple buy-and-hold rule fell below breakeven.

A locked pool is not a safe token. The pool lock protects only the liquidity inside the pool. On the launcher skill described above, up to 90% of supply can be set aside in extras such as vaults and airdrops, leaving as little as 10% in the pool. A vault can unlock after as few as seven days.

Mistakes are permanent. A wrong fee address or a bad setting is written on-chain at launch. One example configuration in the skill’s documentation routes 20% of trading fees to a fixed third-party address. An agent copying that example would send fees there unless someone caught it.

Accountability is unsettled. In the same October 2026 remarks, the CFTC chair asked whether an agent’s developer would become subject to registration, and who is liable if an agent goes rogue. He also floated a “know-your-agent” check. Those are open questions, not rules.

The trust takeaway

Yes, an AI agent can launch its own token. The harder question is whether anyone checked it, and at peaks of 13,000 launches a day, our judgement is that the answer is usually no. If you are looking at an agent-launched token, check these before you trust it:

• Who or what launched it, and who controls the deployer wallet?
• Is the pool’s liquidity locked, until when, and can you verify the lock on-chain?
• How much of the supply sits outside the pool, and when does it unlock?
• Are the creator’s tokens vested over months, or claimable after a few days?
• Where do the trading fees go, and did anyone choose that on purpose?

These are the same checks a careful person runs on any new token. The difference is speed. An agent can launch faster than any human can review.

That is where Team Finance fits. Team Finance provides token locks, vesting, staking and airdrop tools, and more than 40,000 projects have used it. Where this is going: lock and vesting tools that an agent can call as part of its own launch, so liquidity is locked and its allocation is vested in the same flow. None of that is shipped for agents today.