What is dex-trade-agent-kit?

Dex Trade Agent Kit is a toolkit that lets AI assistants like Claude or Cursor trade cryptocurrency on OKX, a crypto exchange, directly from a chat conversation. Instead of switching between an AI tool and the exchange website, you describe what you want done and the AI calls the right function to do it. The project ships as two separate packages. One is an MCP server, a small program that speaks the Model Context Protocol so AI clients such as Claude Desktop, Cursor, and VS Code can talk to it. The other is a command line tool for running the same trading actions from a terminal. Both run locally on your own computer, and your API keys for OKX stay on your machine instead of being sent to any outside service. Together the two packages expose 145 tools split across 11 areas covering the full trading process: checking market prices, placing spot and futures orders, managing algo orders such as stop loss and take profit, checking account balances, using OKX's earn and staking features, running grid or DCA trading bots, reading crypto news, and viewing signals based on what large traders are doing. Many of the market data tools work without any account or API key at all. Safety features include a read only mode that blocks any order placement, the ability to turn on only the modules you need, and a built in rate limiter to avoid hitting exchange limits too fast. The project also bundles ready made skill files that tell AI agent frameworks when to use which trading feature. Getting started requires Node.js version 18 or newer. You install the packages with npm, run a setup wizard to add your OKX API keys, then register the MCP server with whichever AI client you use, such as Claude Desktop or Cursor. From there you can ask the AI to check prices, place trades, or manage your account, and it uses the underlying CLI commands to do so. The project is open source under the MIT license and is aimed at people who want their AI assistant to trade directly on OKX rather than through a web dashboard.