The World Cup is an excellent scenario for observing changes in wallets.
Daily fan discussions about team qualifications, changes in championship odds, etc., become tradable, priced behaviors in prediction markets like Polymarket. Therefore, World Cup prediction events integrated by mainstream Web3 wallets, viewed over a longer timeline, can indeed be considered a lightweight entry point and starting point for users to engage with on-chain interactions (Extended reading: The World Cup Frenzy, Prediction Markets Take the Stage: How Are Polymarkets Tearing Open a Portal for Mass Adoption?).
Simultaneously, another earlier but more imaginative change is worth noting: when AI Agents begin to enter the wallet scenario, the way users interact with the on-chain world is also poised to change.
For instance, in this World Cup-related exploration, imToken started experimenting with integrating AI Agents into actual use cases. Features like the web-based proxy and the one in Discord can help users more naturally complete betting transactions based on specific prediction needs, allowing users to no longer operate solely within the wallet app but to easily participate in prediction markets in scenarios like Discord or web pages, seamlessly guided back on-chain by the Agent.
This could very well be an early form of the Agentic Wallet—the future Web3 wallet might not be limited to a wallet application but will likely be an 'AI wallet form' that is as ubiquitous as possible.
I. The World Cup Experiment: When Agents Start Understanding 'Intent'
Returning to our perennial question, over the past decade, the core problem wallets solved was clear: where assets are stored, who controls the private key, and who signs transactions.
For example, when users open imToken, it's mainly to check balances, perform transfers/interactions, and manage multi-chain assets. Consequently, wallets at this stage resemble asset and signature gateways. It's sufficient for users to know what they plan to do and then use the wallet to complete the final step.
But the change in the Agentic era lies in the fact that users might not initially know what they need to click.
In a World Cup scenario, an average user might not first think, 'I need to open Polymarket, find a specific market, evaluate the odds, and complete the transaction.' They are more likely to say, 'What are the highlights of tonight's match?', 'I'm confident Portugal will qualify; are there relevant markets?', 'Is this betting line already too low?', or 'If I just want to participate with a small amount, what's the process?'
Previously, these questions might have been scattered across WeChat groups, social media discussions, or search engines, requiring users to piece information together and execute step by step. However, with Agent intervention, the interaction method has changed significantly: users only need to express a general intent, and the Agent will proactively break it down into a path, with the wallet then responsible for turning that 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 wallets are starting to evolve from 'function menus' into 'intent interpreters.' Previously, wallets primarily required users to decide whether to transfer, swap, stake, or connect to a DApp? Future wallets might be more straightforward, requiring users to simply tell them in natural language what they want to accomplish.
This is why a mass event like the World Cup is well-suited as an entry point for Agentic Wallets. It naturally provides context, users naturally have the desire to express themselves, and they naturally need to make judgments. The Agent doesn't need to start by managing complex asset portfolios for users, especially those involving intricate asset management with higher risks. It can first help users find interaction paths within a specific scenario before handing final control back to the wallet and the user.
The web-based and Discord-based proxies capable of on-chain interaction mentioned at the beginning of the article, which imToken piloted, are typical examples. They bring wallet capabilities to a lighter-weight entry point, allowing users to find interaction paths through an Agent on an event page, within a World Cup scenario, without necessarily first opening the app or entering the traditional DApp browser.
This indicates that the boundaries of wallets are expanding outward.
Previously, wallet entry points were relatively clear: users opened the app, entered the asset page, clicked functions, and connected to DApps. Future wallet entry points might be scattered across more places, such as web pages, Discord, Telegram, AI chat boxes, event pages, developer tools, or even a lightweight wallet interface generated by the user.
From this perspective, World Cup predictions are not the main point; the key is that they allow wallets, for the first time, to more naturally position themselves ahead of 'user intent.'
II. From Agent Pay: AI is Entering the Payment Layer
If viewed only within the Crypto sphere, Agentic Wallets might still be easily understood as a narrative concept—AI helping users watch markets, find opportunities, and make trades, sounding reminiscent of the last wave of AI Agent hype.
But Mastercard's launch of Agent Pay for Machines on June 10th suddenly made this matter extend beyond Web3.
Mastercard's definition of Agent Pay is clear: enabling trusted AI Agents to participate in payments with user authorization, including how Agents are identified, authorized, verified within the payment network, and how merchants, issuers, and users know a transaction was assisted by an Agent.
This is highly similar to the problems faced by Web3 wallets.
When AI only helps you write copy, the cost of errors is usually manageable. But once AI starts participating in asset interactions, the problem shifts: does it actually have permission? Is its understanding of user intent accurate? Are the services it calls trustworthy? Does the transaction it initiates exceed boundaries? If the outcome doesn't match user expectations, who provides recourse?
Mastercard's answer is to redesign the 'trusted Agent's' identity, tokens, authorization, risk control, and dispute handling within the payment network.
This signal is crucial. If the Agentic concept in the Web3 circle still carries a bit of geeky imagination, the fact that traditional financial giants are starting to design payment infrastructure for Agentic Commerce indicates this has entered a more realistic business context.
Looking closer, domestic payment giants are also moving in this direction. WeChat Pay is testing AI payment features in collaboration with Tencent's intelligent agent product WorkBuddy, such as launching an 'AI Exclusive Card' within the WeChat wallet. Based on disclosed information, the core of such products is not to let AI spend money arbitrarily but to set boundaries for intelligent agent payments through methods like recharge limits, payment authorization scopes, and password verification confirmations.
This logic aligns with Mastercard's Agent Pay: AI can participate in payments, but it must be identified, authorized, restricted, and audited.
Web3 wallets face the on-chain version of the same problem. The difference is that traditional payment systems emphasize the network, merchants, issuers, and compliance responsibilities; on-chain wallets emphasize private keys, signatures, authorizations, contract calls, and user self-custody.
Precisely because of this, Agentic Wallets cannot simply copy the traditional payment path. After all, in traditional payments, users can rely on banks, card networks, merchants, dispute resolution, and risk control systems. But in the on-chain world, once a transaction is on-chain, there's often no undo button.
The higher the efficiency brought by AI Agents, the more critical the last line of security defense borne by the wallet becomes. Therefore, future wallets cannot just let Agents 'do things'; they must ensure Agents 'can only operate within user-allowed boundaries.'
This is also why Web3 might actually be more suitable for discussing Agentic Wallets.
III. The Agentic Era: How Are Wallets Redefining On-Chain Interaction?
When many discuss AI Agents, they easily drift towards an extreme imagination: in the future, AI will automatically trade for me, manage my finances, find airdrops, and arbitrage.
This direction is certainly attractive, but for wallets, the real difficulty is not 'automation' but 'a sense of boundaries.'
Because a wallet is not an ordinary application. In ordinary apps, if AI recommends the wrong song, writes the wrong passage, or clicks the wrong page, it's at most an experience issue. But in a wallet, if AI misunderstands an instruction, calls the wrong contract, or grants excessive permissions, it could become a real asset risk (Extended reading: Sign Isn't Just Signing: When an AI Agent Signs for You, Who Holds the Control?).
Therefore, the primary question for Agentic Wallets is not 'how much can AI do for the user?' but 'how does the user know what AI is doing?'
This is also why imToken has consistently emphasized control over the past few years. From self-custody wallets to multi-chain asset management, to today's AI co-creation and Agent exploration, the truly continuous thread isn't about having more and more features, but about users always being able to understand, confirm, and control their digital world.
In the Agentic era, this thread becomes more concrete.
Wallets need to let users understand who this Agent is, what capabilities it can call, how long the authorization lasts, whether it can operate across DApps, when it requires my reconfirmation, whether I can pause or revoke it with one click... These questions might sound tedious, but they are the foundation upon which Agentic Wallets can truly be established.
Because the power of an AI Agent lies precisely in its ability to simplify complex processes. The user says one thing, and the Agent can break it down into a dozen-step execution path. This is good for experience but also a challenge for security. The longer the path and the more intermediate steps, the more the wallet needs to bring key nodes back into the user's view.
Good wallet interactions in the future might not involve showing users more technical details, but translating complex transactions into language users can understand:
- You are authorizing a specific Agent to call a specific contract within the next 24 hours;
- This operation can use at most a certain amount of USDC;
- It can only access World Cup-related prediction markets, not touch your other assets;
- Each transaction exceeding a certain amount requires reconfirmation;
- Authorization automatically expires after a set period;
- You can pause this Agent anytime within the wallet;
This might sound like a distant future, but it's already starting from some small scenarios. The World Cup prediction event is one event-based entry point, and imToken's pilot web-based Agent and Discord Agent are community-based entry points.
Especially at times like these, wallets cannot retreat to the background.








