Agents · The Agents On-Chain reference

AI agents in crypto: the questions people are asking, answered

AI agents can now hold a wallet, pay for things and trade on their own. Most people's first question is not what an agent can do. It is whether it can be trusted with money. This hub answers the 48 questions people ask most about AI agents in crypto and Web3, in plain English, with the risks named and the sources checked.

Each guide is the written companion to one episode of Agents On-Chain, TrustSwap's explainer series. Start at the beginning if agents are new to you, or jump to the question you came with.

48Questions, one guide per episode
6Series, from basics to what comes next
~176MAgent payments, worth about $73M, in the year to April 2026
3 in 4Online adults uncomfortable letting an agent pay on its own
01 · What it is

What an AI agent is, and why crypto is where it shows up first

A chatbot answers. An agent acts. Give an agent a goal and it uses tools, data and sometimes money to reach it, without someone pressing send at every step. That last part is why agents and crypto keep appearing together.

Software cannot walk into a bank and open an account. It has no ID and no signature. It can control a crypto wallet, though, and pay or be paid in seconds, at any hour, in amounts as small as a few cents. A crypto investment firm counted about 176 million agent transactions worth roughly $73 million in the year to April 2026. That is real activity, and it is still tiny next to the rest of crypto.

The real question is trust, not capability

A major research firm found in 2026 that about three-quarters of online adults in the US, UK and Canada were uncomfortable letting an agent pay on its own, even with spending limits. Crypto users are more open: in a survey of 2,632 crypto users, 87% said they would let an agent manage at least a tenth of their portfolio. Yet only about a third said they would trust AI with wallet access.

That gap is what this hub is about. The concerns come up again and again:

  • Loss of control. What can the agent touch, and what can it do without asking?
  • Errors and liability. When an agent makes a mistake on a blockchain, the action usually cannot be undone. Who pays?
  • Manipulation. Agents can be tricked by hidden instructions, poisoned data or a hijacked dashboard.
  • Hype. Many "agents" are ordinary bots, and many registered agents show no sign of a working service.

The six series

The 48 guides follow the show's six series, in the order a newcomer needs them.

SeriesThe question it answersGuides
1 · What AI agents really areWhat an agent is, what it can do on a blockchain today, and what is still hype1–8
2 · Agents and moneyHow software holds, spends and sends value, and how to limit it9–16
3 · Agents, exchanges and tradingWhat happens when you connect an agent to an exchange account17–24
4 · When it goes wrongReal incidents, how they happened, and the safeguard that would have helped25–32
5 · Rules, liability and responsibilityWhat regulators have said, who is accountable, and what is still unsolved33–40
6 · Tokens, treasuries and what's nextAgents that launch tokens, run treasuries and invest, and where this is going41–48
02 · How to use it

Read it in the order the risk grows

Every guide follows the same shape as its episode: the question, the short answer, how it actually works, what could go wrong, and one takeaway on what to check before you trust an agent with anything. The path below runs from understanding an agent to letting one near your money.

  1. Understand it. Separate agents from bots and hype before anything else. Start with what an AI agent is and what agents can actually do on a blockchain today.
  2. Follow the money. Learn how agents pay and how their spending is capped. See how an agent pays for something and giving an agent a spending limit.
  3. Connect carefully. Before an agent touches an exchange account, decide what its keys allow. See connecting an agent to your exchange and withdrawal permissions.
  4. Learn from failures. Real incidents show where the safeguards were missing. See the $250,000 mistake and prompt injection.
  5. Know who is accountable. Most of the law here is unsettled. See who is responsible when an agent loses money and what the SEC and CFTC have said.
  6. Look ahead. Agents are already launching tokens and managing funds. See agents that launch tokens and what happens when most users are machines.

If you only read three, read how to check whether an agent is legit, whether an agent should ever have withdrawal permissions and whether you can get money back after an agent sends it to the wrong place.

03 · TrustSwap and this subject

Everyone is building agents that can move money. We are asking what they can be trusted with.

TrustSwap builds the infrastructure that makes token launches verifiable: liquidity locks, vesting and the checks a careful buyer runs before trusting a token. Those checks matter more, not less, when the buyer is software. An agent buying a newly launched token needs the same lock and vesting checks a careful human runs, and it has to run them every time.

Most of the guides in this hub never mention a TrustSwap product. Where one genuinely answers the question, the guide says so, and it describes only what the product does today. Agent-facing features are described as where this is going, not as shipped.

Token infrastructureTeam FinanceLiquidity locks, vesting, staking and airdrop tools, used by 40,000+ projects. Relevant to agents that launch or buy tokens.
News and portfolioThe Crypto AppCrypto news and portfolio tracking. Relevant to how agents "see" markets and to apps that move from alerts to actions.
Custody and advisoryPanthea and AlcyoneInstitutional custody, and TrustSwap's strategy and advisory practice. Relevant to agents near treasuries, tokenomics and advice.

The 48 questions

04 · Guides and related hubs

Six series of eight, in episode order. Each guide answers one question in plain English, names the risk honestly and lists what to check.

Related hubs

Platforms

FAQ

Common questions about AI agents in crypto

What is an AI agent in crypto? An AI agent is software that is given a goal and takes actions to reach it, using tools, data and sometimes money. In crypto, that usually means it controls a wallet and can send payments or trades on a blockchain without a person approving every step.

Are AI agents safe to use with my money? Some uses are safer than others. The risk depends on what the agent is allowed to touch: an agent with trade-only access and a spending limit can make bad trades, while one with withdrawal rights can lose everything. Most actions on a blockchain cannot be undone, so limits matter more than promises.

Who is responsible if an AI agent loses money? As of November 2026, the question is officially unsolved. Regulators have raised it publicly, and in practice the terms of most platforms put the risk on the account holder. Read the terms before connecting an agent to any account.

Why do AI agents use crypto instead of bank accounts? An agent cannot open a bank account because it has no legal identity. It can control a crypto wallet, and it can pay or be paid instantly in very small amounts, which card networks cannot do profitably.

Does TrustSwap make AI agents? No. TrustSwap builds token infrastructure, including liquidity locks and vesting through Team Finance. This hub explains how agents work and what to check before trusting one; it does not recommend any agent.

Agents On-Chain

One question per episode.

Watch the series in order, or start with the episode that answers the question you came with.

This hub is educational content only and is not financial, legal or tax advice. Legal and regulatory points summarise public sources as of November 2026, should not be relied on, and may describe rules that are proposed or still coming. Facts current as of November 2026.