Bybit has rolled out an AI trading hub that lets users execute crypto trades through plain-language commands. The new feature, called AI Trading Skill, works through assistants and tools such as ChatGPT, Claude, Gemini, OpenClaw, Cursor, and Windsurf. Instead of using a standard trading dashboard or writing code, users can enter requests like buying BTC at market price or checking portfolio performance, and the system converts those requests into trading actions.
253 API endpoints connect the chat layer to exchange infrastructure
Behind the conversational interface, the product is tied into 253 API endpoints across Bybit’s infrastructure. According to the release, the setup covers six core areas: real-time prices, candlestick data, order book depth, and funding rates; market orders, limit orders, and batch transactions; leveraged positions, stop-loss and take-profit settings, and conditional trades; flexible savings and on-chain earning products; balance checks, deposits, withdrawals, and asset conversion; plus margin borrowing, spread trading, and options RFQ pricing through real-time WebSocket streams.
Bybit presents the product as a lower-friction way to connect AI tools to exchange functions. The company says users only need to install a GitHub skill file, with no package manager setup, interface work, or configuration files required. Once enabled, the AI skill syncs with platform updates automatically so users can access newly added trading functions.
Testnet first, then manual confirmation for live orders
Bybit built several control layers into the rollout. New users are directed to a testnet environment first, where they can try AI-generated commands with simulated funds before moving into live markets. After that switch, every live transaction still requires manual confirmation, keeping the final order decision with the trader rather than the assistant.
The exchange also uses secure API authentication to connect AI assistants to its backend. Instructions produced by the AI are translated into verified API calls and are executed only after passing the platform’s security checks. The stated goal is to reduce the risk of unauthorized instructions and exposure of sensitive account credentials.
Conversational trading expands, but execution risk remains
The launch reflects a wider move from AI tools that only provide market insight toward systems that can carry out actions under user supervision. In that model, AI can help monitor markets around the clock, respond to price moves, and handle multiple positions with less manual switching between tools.
The source also notes that the risks have not disappeared. Poorly written prompts or misread instructions can trigger unintended trades. Fast execution can magnify losses, especially in leveraged derivatives markets. It also points to prompt-injection attacks and phishing attempts as potential threats to AI-driven workflows. The article says analysts recommend starting with small positions and monitoring AI-directed activity closely.