KOSUNI CRYPTO
Korea's Crypto Pulse, in English
$BTC $ETH $XRP $SOL $DOGE
Exchanges
Neutral

MEXC Ventures: AI Trading Agents Shift Focus to Order Execution and Trust Design

Published September 22, 2026 4:39 PM · 16 views $BTC
MEXC Ventures: AI Trading Agents Shift Focus to Order Execution and Trust Design

Crypto exchange artificial intelligence (AI) competition is moving from explaining market data via chatbots to executing orders through agents. MEXC Ventures analyzes in a report that by 2026, major exchanges' AI strategies will diverge into 'open' models connecting external tools and 'closed' systems handling analysis to trading within their own apps.

Key factors differentiating exchange AI competitiveness include expanding functionality alongside order permission scope, user verification procedures, and revocation mechanisms. While 2023 AI focused on price checks and news summaries, by 2026 it will transition to converting natural language into executable trade conditions—such as splitting purchases when Bitcoin (BTC) falls below $80,000.

Supporting this shift is technology like the Model Context Protocol (MCP), a standard for connecting tools. Exchanges provide standardized tools for price checks, account verification, and order execution, allowing external AI to call them. This enables exchanges to broaden AI touchpoints without relying solely on internal chatbots, while users can manage multiple exchange data through familiar tools.

Binance's Agent OS, launched in August 2026, is a developer platform connecting external AI tools to exchange features. It places agents in dedicated sub-accounts separated from main balances and allows individual permission settings or revocation. Within weeks of launch, daily calls exceeded 280,000—a case where design limiting fund and feature access was introduced alongside service expansion.

OKX's open-source Agent Trade Kit offers 83 tools covering spot, swaps, futures, options, and grid bots, running locally with API keys stored on user devices. Kraken CLI features 134 commands and an MCP server supporting simulated trading using real prices. The report notes Kraken requires user confirmation for orders even when AI suggests opportunities.

Binance added AI Trading Skills in March 2026, announcing 253 connected API endpoints. MEXC Ventures identifies the advantage of open exchange AI as expanding data utilization scope—aggregating balances from multiple exchanges on one screen or processing trade history into settlement reports, simplifying procedures previously requiring API documentation and coding.

Exchanges also pursue 'closed' strategies to complete user experience within their apps. MEXC's September 2026 AI upgrade features a three-step process: discovery, understanding, and trading. The AI assistant analyzes real-time prices, holdings, and trade history together, allowing users to ask questions while viewing charts in a chat window. The AI Radar adds financial news visualizations, market briefing videos, and information on US stocks and key events.

MEXC's execution tool, AI Strategy, converts natural language ideas into strategies for user review before automatic execution. MEXC also previews an agent-focused command-line trading interface—interpreted as adding external connectivity to their app-centric AI strategy.

Bitget has rolled out GetAgent since July 2025, followed by developer-focused Agent Hub in February 2026, Telegram-based GetClo in March, and Strategy Library (GetAgent Playbook) in June. The Playbook allows users to select pre-built strategies. Bitget expands into tokenized US stocks, enabling crypto and stock exploration and orders within a single account—a closed strategy focusing on reducing tool installation and integration burdens.

Coinbase concentrates on AI agent payment infrastructure rather than trading interfaces. Its open payment standard x402 uses HTTP 402 status code to design agents paying in stablecoins on-chain. The report cites over 200 million cumulative transactions processed, mostly with USDC, supporting the agent economy by enabling micro-payments for API access and paid data without automating orders.

The boundary between open and closed models is not fixed. As exchanges increasingly operate both internal AI services and external connection tools, distinctions narrow to where resources are focused—connectivity for users processing multiple exchange data versus accessibility for mobile app users analyzing and trading within one interface.

MEXC Ventures identifies 'trust design' as the key turning point in future exchange AI competition. It emphasizes distinguishing between automatic execution after strategy approval versus per-order confirmation, and clarifying how much funds and permissions agents may access. The report states that automation scope, user verification timing, and liability for malfunctions remain insufficiently resolved. By second half of 2026, the ability to clearly control the execution process will become as important as the number of features in AI agent competition.

Korean Source

This article is an English localization of a Korean-language crypto news report. Original headline: “AI에 주문 권한 어디까지 맡길까?”…멕시벤처스(MEXC Ventures), 거래소 AI 경쟁의 핵심은 ‘신뢰 설계’