How AI agents can transform DeFi trading without sacrificing user control

AI agents have moved from experimental tools to active participants in financial markets, and Neyro’s Andrew Isaacs has argued that decentralized finance could become one of the sectors where the technology proves its value most clearly.

Over recent weeks, several major companies have already started rolling out products designed around that same idea. 

Robinhood launched Agentic Trading and Agentic Credit Card services that allow approved AI agents to execute trades and purchases through dedicated accounts with user-defined limits. 

Coinbase-backed Base introduced Base MCP, a system that connects AI assistants such as ChatGPT, Claude, Codex, and Cursor to crypto wallets for tasks ranging from token swaps to portfolio monitoring. 

Meanwhile, Coinbase has expanded its x402 payment infrastructure and agentic commerce initiatives, which CEO Brian Armstrong said could support an economy that eventually exceeds the scale of human commerce.

As AI agents take on more financial responsibilities, some industry participants believe decentralized finance could become one of the most important testing grounds for the technology.

Isaacs, who serves as the Chief Operating Officer at Neyro, a decentralized AI-powered crypto trading platform, believes trading presents conditions that quickly reveal whether AI systems can perform reliably when decisions carry real financial consequences.

“In many industries, an AI agent can save time. In trading, it can show whether this new model of automation is actually reliable under pressure,” Isaacs said in comments shared with crypto.news.

Unlike routine business workflows, trading requires continuous monitoring of market activity, interpretation of incoming data, and decision-making within predefined limits. Isaacs said those characteristics make markets a useful environment for evaluating how well AI agents operate outside controlled demonstrations.

“Trading is where small delays and bad judgment show up immediately,” Isaacs said. “That makes it a very honest environment for testing what AI agents can actually do.”

Fast-moving DeFi markets create a case for automation

Business interest in AI agents has already moved beyond experimentation. A McKinsey survey found that nearly two-thirds of companies are testing AI agents in their operations, while crypto firms have increasingly focused on ways to connect those systems with financial infrastructure.

For decentralized finance, Isaacs said the opportunity comes from the speed and complexity of crypto markets themselves.

“What stood out to me was the mismatch between how fast DeFi markets move and how manually most users still operate,” he said.

Around-the-clock trading, fragmented liquidity, and thousands of tokens spread across multiple blockchains have created an environment where keeping up with every opportunity is difficult for individual traders.

“A human trader cannot watch every pool, every token. But an AI agent can. That creates a very obvious use case for agentic systems,” Isaacs said.

Recent developments across the industry point to the same direction. Base MCP was introduced to help users manage crypto activity through AI chat interfaces while requiring transaction approvals before execution. 

Similarly, Isaacs’s Neyro is attempting to combine automation with decentralized infrastructure, allowing users to benefit from AI-driven trading without relying on custodial systems.

Coinbase has also highlighted the growing use of USDC and Base for machine-to-machine payments, stating during its first-quarter earnings call that AI agents use USDC in 99% of tracked transactions and conduct more than 90% of those payments on Base.

According to Coinbase, AI agents are already using x402 infrastructure for services such as trading, data access, AI inference, media generation, and storage. Official x402 statistics cited by the company showed monthly volumes surpassing 75 million transactions.

Why decentralized exchanges have lagged behind

Despite growing interest in AI-powered finance, Isaacs said decentralized exchanges have faced obstacles that centralized platforms do not.

Centralized trading environments allow developers to deploy AI systems within platforms that already have internal controls and safeguards. On decentralized exchanges, however, agents interact with non-custodial wallets and smart contracts where transactions are typically irreversible.

“There was also a trust problem. In CeFi, an AI system can sit behind an exchange account with internal controls. But on DEXs, AI touches wallets and smart contracts directly. A bad prompt or a misread market condition can become an irreversible transaction,” Isaacs said.

Security concerns surrounding AI agents have continued to attract attention across the industry. When Base launched MCP, the company emphasized that transactions require explicit user approval and that the system never accesses private keys. 

Separately, researchers from organizations including Google, Meta, Gray Swan AI, EmbraceTheRed, and several universities argued in a recent report that AI agents should be treated as untrusted components and isolated from sensitive instructions and data wherever possible.

Against that backdrop, Isaacs said the industry should be careful not to sacrifice decentralization in pursuit of convenience.

“AI is powerful enough to make centralization look convenient again, and that is the danger. The point of Web3 was never only to make financial products more digital. It was to change the trust model.”