Korbit Unveils AI Agent Trading Tool Tied to ChatGPT, Claude
Summary
- Korbit said it launched Korbit CLI, an installable program that supports generative AI and AI agent trading.
- The company said it will also provide a paper-trading environment, testing based on the Korbit open API, Korbit TUI for terminals and llms.txt.
- CTO Lee Jung-woo said the MCP server, agent skills, paper-trading environment and llms.txt provide a foundation for AI to use Korbit's functions accurately and safely.
Forecast Trend Report by Period
South Korean cryptocurrency exchange Korbit has unveiled an AI agent trading environment that uses generative AI tools such as ChatGPT and Claude for tasks ranging from price checks to order execution.
Korbit said July 24 that it had launched an installable program called Korbit CLI, short for Command Line Interface. The tool is installed on a user's PC and integrates with AI agents including ChatGPT Codex and Claude Code, allowing users to access exchange functions through natural-language commands. Those functions include checking prices and balances, placing orders, and making deposits and withdrawals.
Korbit CLI includes an MCP, or Model Context Protocol, server and AI agent skills. It was designed to let AI agents understand and use Korbit's functions and carry out user requests without requiring a separate development environment.
The company also offers a paper-trading environment that lets users test AI-based automated trading programs without using real assets. Developers can test orders, executions and balance changes in the same environment as Korbit's open API and verify how their programs behave.
Korbit is also offering Korbit TUI, or Text-based User Interface, for developers who prefer terminal-based environments. Users can view real-time prices, order books and candlestick charts on a terminal screen and use key functions including trading, deposits and withdrawals.
The exchange also released an AI-friendly document called llms.txt to make its open API documentation easier for AI systems to understand. Developers can provide the file to AI coding agents to build price-collection programs and trading bots more efficiently, Korbit said.
Lee Jung-woo, Korbit's chief technology officer, said that in environments where AI handles real assets, reliability, traceability and a sufficient pre-validation framework matter alongside model performance. He said the MCP server, agent skills, paper-trading environment and llms.txt provide a foundation for AI to use Korbit's functions more accurately and safely. Korbit plans to keep expanding infrastructure that allows AI agents to connect directly and trade cryptocurrencies, he added.
