Binance, the world's largest cryptocurrency exchange, has launched Agent OS, a platform allowing AI agents to analyse markets and execute trades on users' behalf, marking a significant step in bringing autonomous AI into real-money financial management. The move reflects a broader shift in the AI industry from passive chatbots to agents capable of taking direct action, but Binance is placing much of the responsibility for controlling these agents onto users themselves, who must configure access permissions and set trading limits.

Agent OS connects developers' AI applications to Binance's existing infrastructure, including its APIs, Wallet Agentic Hub and Skill Hub, and now supports the Model Context Protocol, working with tools such as ChatGPT, Codex, Claude Code and Cursor. Control is enforced primarily through "sub-accounts" that users assign to agents, with withdrawals blocked by default and no separate cap on trading losses beyond the funds transferred in. Binance vice president Jeff Li said the exchange cannot see an agent's underlying reasoning, as this occurs within the user's own AI application, meaning it can monitor trading activity but has limited insight into whether decisions were influenced by faulty data or manipulation, such as prompt-injection attacks.