Coinbase launches tool for AI agents to manage trading and payments

What happened?

Coinbase has launched Coinbase for Agents, a tool that allows artificial intelligence agents — such as ChatGPT, Claude, or others — to execute cryptocurrency trading operations and manage payments on behalf of users. According to CNBC, users can instruct their agent to rebalance portfolios, identify trading opportunities, execute strategies, and manage positions over time. The company plans to expand these capabilities to stocks and predictions in the future. The tool relies on the machine-to-machine payment protocol x402, created by Coinbase in May 2025. This protocol has already processed over 100 million transactions, and according to data from x402scan.com, approximately 157,000 agents have acted as buyers in the last 30 days. This announcement marks a milestone in the evolution of AI agents from simple assistants to autonomous economic actors, a trend that has been gaining momentum since the emergence of the first financial chatbots a decade ago.

Why is it important?

This launch represents a paradigm shift: AI agents move from being mere assistants to becoming autonomous economic actors. Lincoln Murr, AI product lead at Coinbase, explains:

“The idea is to give agents access to money and, through that financial independence, enhance their capabilities for virtually anything on the internet. In the 2010s, every internet company faced the transition from desktop and web to mobile. Now, in the late 2020s, we are seeing exactly the same thing: agents will be the new primary economic actors on the internet.”

The ability of agents to pay for digital services — such as research behind paywalls, data APIs, or on-demand computing — without human intervention eliminates the need to manage logins or traditional subscriptions. This paves the way for agentic shopping, where agents will search, compare prices, select, and purchase products on behalf of users. The potential impact is enormous: according to a MarketsandMarkets report, the autonomous agent market could reach $28.5 billion by 2030, with a compound annual growth rate of 34%.

Consequences and context

Coinbase's move is part of a broader trend toward automating financial transactions using AI. Major tech companies like OpenAI and Anthropic have already shown interest in autonomous payment capabilities for their models. However, direct integration with a cryptocurrency exchange platform gives Coinbase a significant competitive advantage. Historically, the financial industry has been cautious about full automation; recall the 2010 flash crash, when high-frequency trading algorithms caused a sudden market drop. But unlike those algorithms, today's agents can learn and adapt, posing new challenges. From a regulatory standpoint, this tool raises questions about responsibility for transactions carried out by autonomous agents. Who is liable if an agent executes an unwanted trade? Coinbase has not yet detailed the control mechanisms or limits users can set. In the United States, the SEC is already examining the use of AI in finance, and in the European Union, the AI Act classifies autonomous trading systems as high-risk, requiring human oversight. For businesses, this presents an opportunity to reduce operational costs and streamline payment processes. For example, a company could delegate the purchase of software licenses or data subscriptions to an agent, saving time and human errors. For individual users, the promise of delegated financial management may be appealing, but it also carries security and privacy risks. If an agent is compromised, an attacker could gain access to funds. Coinbase assures that agents only act within predefined limits, but experience shows that no system is infallible.

What should readers know?

• Access: The tool is available to Coinbase users who integrate their accounts with compatible AI agents. A verified account and specific permission settings are required.
• x402 Protocol: It is the engine for autonomous payments; it has demonstrated scalability with over 100 million transactions since its launch in May 2025. According to x402scan.com, there have been about 157,000 active buyer agents in the last 30 days, indicating significant early adoption.
• Security: Coinbase emphasizes that agents can only act within the limits set by the user, but full autonomy remains a risk. The company recommends setting daily spending limits and periodically reviewing transactions. Additionally, agents operate on Coinbase's own Base blockchain, providing transparency but also immutability of transactions.
• Future: The tool is expected to expand to stocks and predictions, and to inspire other platforms to offer similar functionalities. Companies like Stripe and PayPal are already exploring automated payments with AI, but integration with crypto assets gives Coinbase an edge in speed and transaction costs. In the long term, we could see agents negotiating with each other, creating an autonomous machine economy.

In summary, Coinbase for Agents is a concrete step toward an economy where AI agents not only recommend but execute. The question now is how much control we are willing to cede. The history of financial automation teaches us that convenience often comes with new risks. As Murr said, we are facing a change similar to the shift from desktop to mobile, but this time the actors are not humans but algorithms. The key will be to design robust safeguards that allow us to harness the potential without exposing users to unnecessary dangers.