In the past five days, OKX, BingX, and Kraken, three of the world’s largest Bitcoin and cryptocurrency exchanges, have launched new infrastructure designed to allow artificial intelligence (AI) agents to operate autonomously on their platforms.
The three releases are pointing in the same direction. This means moving from interfaces designed for humans to machine-readable tools. AI agents can read the market, make decisions, and execute orders without human intervention.
This news comes in the context of former Binance CEO Changpeng Zhao (CZ) and Coinbase CEO Brian Armstrong recently stating: AI agents will dominate the number of transactions and payments made by humans And virtual currencies will become the natural financial infrastructure for this.
OKX: 82 tools, open source, keys that never leave your device
On March 10th, OKX released the Agent Trade Kit. It is a set of 82 tools organized into 7 modules that cover the entire trading cycle, from market data to order management, positions, trading bots, and account auditing.
Agent Trade Kit is designed for developers building agents. Not intended for retail users. The most obvious differentiator is privacy. Your API (Application Programming Interface) key is stored only on your device, with no external servers or data ever leaving your device.
The Agent Trade Kit code is open source under the MIT license, so Can be freely audited or modified.
BingX: Natural language and USD 300 million in bets for traders
BingX also launched an AI Skills Hub on March 10: covering 15 skill modules Perpetual futures, spot trading, account managementwhich can be manipulated using natural language.
Unlike OKX and Kraken, BingX is explicitly aimed at lowering the barrier to entry for regular traders. Instead of writing code or commands, similar to everyday use of ChatGPT, users write what they want to do and the agent does it.
The launch is part of what the company calls BingX AI, a $300 million commitment aimed at building “the first AI-native exchange.”
Kraken: 134 commands for working with agents without a graphical interface
On March 6th, Kraken introduced a command line interface (CLI) tool. Rather than opening like an application with buttons and menus, Operate by writing instructions in text From a command window, the same environment that developers use to run their code.
Kraken’s new tools work right out of the box, without the need to install any additional programs or configure a server. On top of that, Contains 134 commands aimed at trading operations: spot, futures, funding rates, paper trading (simulation operations without real money).
134 commands return the information required for trading (price, active orders, balance, errors) in a structured format. Can be read by external AI agents and directly interpreted.
External agents built by developers use the Kraken CLI as a connection channel with this exchange. Send instructions through commands and receive response data to make next decisions.
In this way, agents can explore the market, evaluate information, and execute orders autonomously without human intervention at each step. Kraken’s release is aimed at developers building AI agents, not general users.
As a demonstration, Kraken provided access to an agent in paper trading mode that simultaneously manipulates 16 indicators of crypto assets such as BTC, ETH, and SOL. That is, we simulated the operation without using real money before moving on to the actual trading.
Coinbase and Binance accelerate AI adoption on exchanges
The world’s largest Bitcoin and cryptocurrency exchanges, such as Binance and Coinbase, are leading the trend of these platforms developing AI technology.
In August 2025, Coinbase launched the x402 protocol, a payments infrastructure that allows AI agents to automatically settle transactions with each other, and in February 2026 added Agenttic Wallet, a wallet designed to allow agents to maintain and manage funds without human intervention.
According to on-chain data, in October 2025, x402 recorded 10,000 transactions and by February 2026, it exceeded 100 million transactions. 999.900% increase.
Binance jointly developed Binance Pay’s agent payment protocol with Google in October 2025, integrated AI tools into the wallet in January 2026, and on March 7th launched AI Agent Skills, a set of seven tools for autonomous agents to execute orders, audit contracts, and track market movements.
What OKX, BingX, and Kraken have done this week is join the race that also includes the two largest exchanges.
Risks in automating financial operations
Delegating financial decisions to autonomous systems means algorithmic errors, technical failures, command injection attacks, or market manipulation. losses may occur Humans cannot interfere with time.
A specific case already documented by CriptoNoticias illustrates this risk. At the end of February last year, an AI agent called Lobster Wild transferred tokens worth USD 617,000 to an unknown person who asked for help in X after interpreting a message requesting the equivalent of 4 Solana Coins (SOL) for medical treatment.
In the same month, decentralized finance (DeFi) protocol Moonwell Loss of USD 1.7 million due to error in AI-generated smart contract: The system priced the asset at $1.12 when its actual value exceeded $2,200. Attackers quickly exploited that difference. This bug passed all human reviews undetected.
Structural issues lie not only in operability but also in design. A report published in late January by Argentinian company Lambda Class warns that Ethereum’s libraries and infrastructure are built on the assumption that transactions will be initiated and understood by people.
When an AI agent signs a transaction without prior human review, the error is no longer theoretical and becomes an irreversible loss.
(Tag Translation) Bitcoin (BTC)
