Korbit Launches AI Agent Trading Tool, Enabling Crypto Orders via ChatGPT — BigGo Finance

South Korean cryptocurrency exchange Korbit made a dramatic announcement on July 24, unveiling an AI agent trading environment that enables everything from price inquiries to order execution and fund transfers simply by conversing with generative artificial intelligence. The era where users can instruct AI assistants like ChatGPT or Claude in natural language—without any programming expertise—to directly execute exchange functions has officially begun.

The installable program launched that day, called ‘Korbit CLI (Command Line Interface),’ is a tool installed on the user’s PC that integrates with AI agents such as ChatGPT Codex and Claude Code. Users can execute core exchange functions using everyday language commands like “Tell me the current Bitcoin price,” “Check my Ethereum balance,” or “Buy 100 XRP at market price.”

Korbit CLI can be installed with a single command line and comes configured with an MCP (Model Context Protocol) server and AI agent skills essential for AI integration. A key feature is its design that allows AI agents to independently understand Korbit’s functions and perform actual tasks based on user requests without requiring a separate development environment setup.

The company also announced it is providing a simulated trading environment where users can safely verify AI-based automated trading programs before committing real assets. In this simulation system, users can test order placement, execution, and balance changes in an environment identical to Korbit’s Open API, allowing developers and traders to pre-check the entire operation of their programs.

For developers who prefer terminal environments, Korbit also introduced ‘Korbit TUI (Text-based User Interface).’ This text-based interface allows users to view real-time prices, order books, and candlestick charts directly on the terminal screen while accessing key functions such as order placement and fund transfers.

Korbit also released ‘llms.txt,’ an AI-friendly document format designed to help AI models understand Open API documentation more easily. By simply passing the file’s address to an AI coding agent, developers can build price collection programs or trading bots far more efficiently.

Lee Jung-woo, Chief Technology Officer at Korbit, emphasized: “In environments where AI handles real assets, model performance alone isn’t enough—execution reliability, traceability, and a robust pre-verification system are critical. Through the MCP server, agent skills, simulated trading environment, and llms.txt, we’ve established a foundation for AI to utilize Korbit’s functions more accurately and safely.” He added, “We plan to continuously expand the infrastructure that allows AI agents to directly connect and trade virtual assets going forward.”

Korbit’s move exemplifies how the global AI agent trend is rapidly expanding into finance and virtual asset trading. Unlike traditional automated trading bots that operate according to fixed algorithms, agent trading powered by generative AI is differentiated by its ability to contextually understand user intent and respond flexibly. Notably, by adopting standardized protocols like the MCP server, Korbit has also opened up possibilities for future integration with various AI models and third-party tools.

The virtual asset industry expects this tool launch to significantly improve exchange accessibility not only for developers but also for general users. However, concerns have been raised about security and error risks associated with AI managing real assets. In response, Korbit maintains its position that these risks will be controlled through pre-verification via the simulated trading environment and a traceable execution structure.