Author: imToken

The World Cup is a very suitable scene to observe changes in wallets.

What fans discuss every day, such as whether the team will qualify or not, changes in championship odds, etc., have become priced transactions in prediction markets like Polymarket. Therefore, the World Cup prediction activities integrated with Web3 mainstream wallets can indeed be regarded as a lightweight entrance and starting point for users to interact on the chain when viewed on a longer time line.

At the same time, another earlier but more imaginative change this time is also worthy of attention, that is, when AI Agents begin to enter the wallet scene, the way users interact with the world on the chain may also change.

For example, during the World Cup-related exploration, imToken began to try to put AI Agent into actual usage scenarios. For example, the launched web page and agents in Discord can help users complete betting transactions more naturally around specific prediction needs, allowing users to no longer only complete operations in the wallet app, but to easily participate in the prediction market in Discord, web pages and other scenarios, and be guided back to the chain by the Agent smoothly and seamlessly.

This may not be the early appearance of Agentic Wallet - the future Web3 wallet may not be limited to wallet applications, but it must be an "AI wallet form" that is as ubiquitous as possible.

1. World Cup Experiment: When Agent Begins to Understand “Intention”

It is still a common question we talk about. In the past ten years, the core problem solved by wallets is very clear, which is where the assets exist, who controls the private key, and who signs the transaction.

For example, when a user opens imToken, it is mainly to check the balance, operate transfer/interaction and manage multi-chain assets. Because of this, the wallet at this stage is more like an asset entrance and a signature entrance. As long as the user knows what he plans to do, and then completes the last step through the wallet, that is enough.

But the change in the Agentic era is that users may not necessarily know what they want from the beginning.

For example, in the World Cup scenario, an ordinary user may not first think of "I want to open Polymarket, find a certain market, judge the handicap, and complete the transaction." He is more likely to say "What is interesting about tonight's game?", "I am optimistic about Portugal qualifying, are there any related markets?", "Are the odds too low?", "If I only participate in a small amount, how will the process go?"

In the past, these problems may have been scattered in WeChat groups, social platform discussions, or search engines. Each user had to put the information together and execute it step by step. However, after the Agent intervened, the interaction method changed significantly: The user only needs to express a general intention, and the Agent will actively help him disassemble the path, and the wallet is responsible for turning the path into a series of on-chain actions.

So this is definitely not as simple as adding a chat box to the wallet.

The real change is that the wallet begins to change from a "function menu" to an "intent interpreter". In previous wallets, users mainly decided whether to transfer, swap, pledge, or connect to DApp? The wallet of the future may go one step further and only need to use natural language to tell it what it wants to accomplish.

This is why mass events like the World Cup are very suitable as entrances to Agentic Wallet. Because it naturally has context, users naturally have a desire to express themselves and naturally need to make judgments. Agent does not need to manage complex asset portfolios for users right from the start. Especially when it comes to complex asset management, the risks are higher. It can help users find the interaction path in a specific scenario first, and then return the final control to the wallet and the user.

The imToken interactive agent on the web and Discord mentioned in the article is a typical example. They bring wallet capabilities to a lighter entrance, so that users do not necessarily need to open the App first, or enter the traditional DApp browser first, but can find the interaction path through the Agent in an event page or a World Cup scene.

This means that the wallet's boundaries are overflowing.

The previous wallet entrance was relatively clear. Users opened the app, entered the asset page, clicked on the function, and then connected to the DApp. In the future, wallet entrances may be scattered in more places, such as web pages, Discord, Telegram, AI chat boxes, event pages, developer tools, or even a lightweight wallet interface generated by a user.

From this perspective, this is why we feel that World Cup prediction is not the point. The point is that it allows the wallet to stand in front of "user intention" more naturally for the first time.

2. From the perspective of Agent Pay, AI is entering the payment layer

If you only look at the inside of Crypto, Agentic Wallet may be easily understood as a narrative concept, that is, AI helps users read the market, find opportunities, and make transactions. It sounds very much like an extension of the previous round of AI Agent craze.

But Mastercard’s launch of Agent Pay for Machines on June 10 suddenly made this matter more than just Web3.

Mastercard's definition of Agent Pay is very clear, which is to allow trusted AI Agents to participate in payments with user authorization, including how the Agent is identified, how to be authorized, how to be verified in the payment network, and how to let merchants, card issuers, and users know that the transaction is assisted by the Agent.

This is actually very similar to the problem faced by Web3 wallets.

When AI is just writing copy for you, the cost of mistakes is usually controllable. But once AI starts to participate in asset interaction, the question changes. Does it have authority? Does it understand user intent accurately? Are the services it calls trustworthy? Does it initiate a transaction out of bounds? If the results are inconsistent with user expectations, who will take responsibility?

The answer given by Mastercard is to redesign the identity, token, authorization, risk control and dispute handling of "trusted agents" in the payment network.

This signal is very important. If Agentic in the Web3 circle still has a bit of geeky imagination, then the fact that traditional financial giants have begun to design payment infrastructure for Agentic Commerce shows that this matter has entered a more realistic business context.

Looking closer, domestic payment giants are also moving in this direction. WeChat Pay is testing AI payment functions in conjunction with Tencent's intelligent product WorkBuddy, such as launching an "AI exclusive card" in WeChat wallet. Judging from the disclosed information, the core of this type of product is not to allow AI to spend money at will, but to set boundaries for intelligent payment through recharge limits, payment authorization ranges, password verification, etc.

This is actually the same logic as Mastercard Agent Pay, that is AI can participate in payment, but it must be identified, authorized, restricted, and audited.

The Web3 wallet faces an on-chain version of the same question. The difference is that traditional payment systems place more emphasis on the network, merchants, card issuers and compliance responsibilities; on-chain wallets place more emphasis on private keys, signatures, authorization, contract calls and user self-hosting.

Because of this, Agentic Wallet cannot simply copy the traditional payment path. After all, in traditional payment, users can rely on banks, card organizations, merchants, dispute resolution and risk control systems, but in the on-chain world, once a transaction is put on the chain, there is often no regret button.

The higher the efficiency brought by AI Agent, the more important the last security boundary assumed by the wallet is. Therefore, future wallets should not only allow Agents to "be able to do things", but also allow Agents to "can only do things within the scope allowed by the user."

This is why Web3 may be more suitable for discussing Agentic Wallet.

3. In the Agentic era, how does the wallet redefine on-chain interaction?

When many people discuss AI Agent, they tend to subconsciously go to an extreme imagination. In the future, AI will automatically help me trade, automatically help me manage finances, automatically help me find airdrops, and automatically help me arbitrage.

This direction is certainly attractive, but for wallets, the real difficulty is not "automation", but "sense of boundaries."

Because the wallet is not an ordinary application. In ordinary applications, if AI recommends the wrong song, writes the wrong paragraph, or clicks on the wrong page, it is just an experience problem at most; but in the wallet, if the AI ​​misunderstands an instruction, calls the wrong contract, or gives too much authorization, it may become a real asset risk.

So the first question of Agentic Wallet is not "how much can AI do for users", but "how do users know what AI is doing".

This is why imToken has been emphasizing control in the past few years. From self-hosted wallets to multi-chain asset management, to today’s AI co-creation and Agent exploration, the real continuous thread is not that there are more and more functions, but that users must always be able to understand, confirm and control their own digital world.

In the Agentic era, this main line will become more concrete.

The wallet needs to allow users to understand who the Agent is, what capabilities it can call, how long the authorization lasts, whether it can operate across DApps, when I need to confirm again, and whether I can pause or revoke it with one click... These questions sound a bit cumbersome, but they are the basis for whether Agentic Wallet can truly be established.

Because the power of AI Agent comes precisely from its ability to simplify complex processes. The user said that the Agent can be broken down into more than a dozen execution paths, which is a good thing for the experience, but it is also a challenge for security. The longer the path, the more intermediate links there are, and the more wallets are needed to bring key nodes back to users.

A good wallet interaction in the future may not be about allowing users to see more technical details, but about translating complex transactions into words that users can understand:

• You are authorizing an Agent to call a contract within the next 24 hours;

• The maximum amount of USDC that can be used for this operation;

• It can only access the World Cup-related prediction market and cannot touch your other assets;

• Every time a transaction exceeds a certain limit, it must be reconfirmed;

• The authorization will automatically expire after expiration;

• You can pause this Agent in your wallet at any time;

This sounds like a very far future, but it actually starts with some very small scenes. The World Cup prediction activity is an event entrance, while the web Agent and Discord Agent that imToken is testing are a community entrance.

At this time, the wallet cannot retreat behind the scenes.