Now built for AI

The BitMEX AI Agent Toolkit is a developer package that lets AI agents trade crypto derivatives on BitMEX natively. The toolkit bundles four components: a native CLI binary, an MCP server for Claude and Cursor, a library of pre-built trading skills, and the underlying REST and WebSocket trading API. Agents connect once and gain full market access with risk controls built in. Your AI calls BitMEX the way it calls any other native tool, with typed parameters and structured responses instead of raw HTTP.

A single shell command installs the complete toolkit. Run brew install bitmex/tap/bitmex-cli with your AI agent. Claude Code, Cursor, OpenClaw, Codex, and others are detected automatically. The installer deploys the CLI binary plus all 49 Claude trading skills into the right directories. No Node.js, Python, or Docker required. The binary is self-contained. Once installed, run bitmex auth set to authenticate with your API key, then bitmex skills list to browse the full catalogue. Full install takes under 60 seconds on macOS and Linux, and the skills are immediately callable from your AI.

The toolkit works with Claude, Cursor, and any MCP-compatible AI client. Claude Desktop, Claude Code, Cursor, Codex, Gemini CLI, GitHub Copilot, Goose, and OpenClaw work out of the box. Agents call BitMEX functions as native tools with typed parameters and structured responses. No REST documentation parsing, no curl command crafting. Integration is zero-config for supported clients — the installer registers the MCP server in the right place automatically. For any other MCP-compatible agent, a three-line JSON config is all that stands between your AI and the full BitMEX trading interface.

Every command and every strategy supports the --testnet flag. BitMEX testnet uses real market data with fake capital. Your AI agent rehearses the full execution path — order placement, fills, margin, funding, liquidation logic — with zero financial risk. The promotion path from testnet to live is built into the testnet-to-live skill, which walks the agent through a validation checklist before touching real capital. No separate testnet configuration, no different API endpoint, no credential juggling. The same CLI flag flips the whole toolkit into a fully simulated environment. Start on testnet, graduate when ready.

Claude skills are structured instruction packages that teach an AI agent how to perform specific tasks. The BitMEX AI Agent Toolkit ships 49 pre-built skills organised into five categories: Core, Strategy, Operations, Recipes, and Integration. Each skill is a Markdown file with YAML frontmatter defining the trigger, inputs, execution flow, and exit conditions. When you ask Claude to run a TWAP, the twap-execution skill activates and brings its full runbook into context. Skills work identically across Claude Desktop, Claude Code, Cursor, and any MCP-compatible client. Every skill is open source on GitHub, so any developer can inspect, fork, or contribute.

MCP and CLI are two interfaces to the same BitMEX trading operations, built into one binary. The CLI is for humans and shell scripts. Commands such as bitmex order new and bitmex position list print formatted output and return standard exit codes. The MCP server is for AI agents. Run bitmex mcp start and the same operations become typed tool calls over stdio, consumable by Claude, Cursor, Codex, and Copilot. A developer writing a cron job for daily rebalancing uses the CLI. An agent planning a strategy in Claude Code uses MCP. Both paths hit the same authenticated client, the same rate-limit budget, and the same testnet and live environments.

The BitMEX AI Agent Toolkit is a collection of servers and toolkits intended to facilitate automated or AI-assisted trading, and does not constitute financial, investment, legal, or tax advice. AI-generated actions may trigger real trades and result in significant losses due to model errors, hallucinations, inaccurate or outdated information, latency, market volatility, slippage, liquidity constraints, technical failures, incorrect parameters, or service disruptions.

The toolkit is provided “as is” and “as available,” with no warranties regarding accuracy, completeness, execution reliability, or continuous availability. The toolkit may rely on third-party large language models (LLMs) that are not owned, operated, controlled, verified, or endorsed by BitMEX; BitMEX disclaims all responsibility and liability for any third-party LLM services or outputs, and you use them entirely at your own risk. You are solely responsible for independently verifying all information, configuring and supervising automated strategies safeguarding API credentials, applying least-privilege permissions (including avoiding withdrawal permissions unless strictly necessary), binding trusted IPs, and testing with small amounts before scaling.

To the maximum extent permitted by law, BitMEX disclaims liability for any direct or indirect losses arising from use of the BitMEX AI Agent Toolkit, and you agree to indemnify and hold BitMEX harmless from any claims related to your use, trading activities, or reliance on AI-generated outputs.

Always seek independent financial counsel and ensure you fully understand the risks before trading. Please refer to our Terms of Service and Risk Disclosure Statement for full details.