Binance Launches AI Trading with Emphasis on User Control

On Thursday, Binance, the premier global cryptocurrency exchange with over 300 million registered users, launched a platform aimed at enabling AI agents to analyze market trends and conduct trades on behalf of users, thus integrating autonomous AI into financial management.

Named Agent OS, this innovative platform allows developers to connect AI applications and agents to Binance’s financial ecosystem. It combines the exchange’s existing tools and services, such as Binance APIs, Binance Wallet Agentic Hub, Binance x402 transaction verification and payment facilitator API, and Binance Skill Hub, while also supporting its Model Context Protocol (MCP). The platform is compatible with technologies like OpenAI’s ChatGPT and Codex, Anthropic’s Claude Code, and Cursor, allowing users to give agents access to market data, review account information, and perform trades.

As AI technology transitions from fundamental chatbots to more proactive agents, Binance emphasizes the necessity for user supervision of these agents. This includes defining the parameters for the agents’ access and trading actions.

“Rather than allowing complete autonomy, we equip users with precise access controls regarding what the agent can do,” stated Jeff Li, Binance’s vice president of product, in a recent interview. “This control is implemented at the account level to protect users’ assets.”

To facilitate this, Binance provides dedicated “sub-accounts” for agents, which users can customize for specific tasks, including spot or futures trading. Withdrawals from these sub-accounts are strictly limited to ensure a secure environment for the agent’s activities, as clarified by Li to TechCrunch.

Users can choose whether a trading AI agent must get approval before each transaction or if it can autonomously execute trades once it has been granted permission, according to a Binance spokesperson. Notably, Binance does not impose restrictions on the trading volume or potential losses of an AI agent; instead, the user’s transfer to the sub-account acts as the cap.

When asked if Binance monitors the reasoning behind an agent’s trading decisions, Li mentioned that such reasoning occurs outside of its systems, on the user’s device or within the chosen AI program. “We cannot determine the logic behind the user’s decisions,” he noted.

While Binance can observe the trading performance of an agent, it lacks complete insight into whether decisions were swayed by false information or manipulation.

Li underscored the importance of sub-accounts as protective measures in the context of potential repercussions if an agent were compromised by a prompt-injection attack or a similar risk. Binance assured that its established security protocols, risk management strategies, and anti-money laundering measures for sub-account APIs will be implemented within Agent OS from the beginning.

Trading remains a central function that Binance seeks to support. However, Li indicated that agents could also analyze markets, conduct research and risk evaluations, respond to signals, and autonomously execute orders or strategies like arbitrage.

Agent OS aims to forge seamless connections between agents and payments, in addition to on-chain activities. With Binance’s x402 integration, agents can manage payments, and the Agentic Wallet permits interaction with tokens and decentralized finance protocols.

Unlike exchange trading, where Binance does not set specific limits on the volume an agent can trade or lose within its sub-account, transactions through the Agentic Wallet are constrained by daily limits established by Binance. Standard swaps are capped at $50,000 per day, DeFi transactions are limited to $100,000 daily, and x402 payments cannot exceed $20 per day, according to the company.

Li characterized Agent OS as Binance’s “first step” towards providing developers with a platform for creating AI-driven applications capable of functioning across both cryptocurrency and traditional financial markets.

Binance is not the only player enhancing its infrastructure for AI agents. Other competing cryptocurrency exchanges are also progressing in this arena, utilizing MCP and additional developer tools to enable AI applications direct access to market data and trading systems.

In March, Kraken introduced an open-source command-line tool featuring a built-in MCP server, enabling AI agents to perform actions such as spot and futures trades. In June, Coinbase launched Coinbase for Agents, which allows AI agents to connect directly to user accounts, execute trades, make payments, and carry out other financial tasks within user-defined parameters. Similarly, OKX has promoted agent-based trading on its platform by releasing an open-source MCP toolbox earlier this year.