AI Agent Integration

Two integration options are available. Use either one, or configure both.

	Option 1: Install skills	Option 2: Configure MCP
Setup	Run one command	Edit the agent configuration file
Invocation path	The agent runs mexc commands in the terminal	The agent calls directly, without going through the terminal
Prerequisites	The agent can execute terminal commands	The agent supports MCP

Option 1: Install Skills

mexc skills install                   # Install for Claude Code by default
mexc skills install --agent cursor    # Install for a specific agent
mexc skills install --all-agents      # Install for all detected agents
mexc skills update                    # Re-run after upgrading the CLI

This installs 25 capability reference documents. After installation, you can ask questions directly, such as "What is the current BTC contract funding rate?" or "What is my spot USDT balance?"

Each document includes a full capability table: command syntax, read/write behavior, whether a key is required, whether repeated execution is safe, and a description. It also includes practical workflows, such as how to confirm an order fill after placing a trade. For questions like "What can spot trading do?" or "Which command shows my positions?", the agent answers from local documents that you can verify.

The documents ship with the binary and match the installed version. After upgrading the CLI, run mexc skills update to sync them. If versions do not match, mexc doctor reports a warning.

Capability Coverage

Use the table below to confirm category coverage. For individual capabilities, query through the agent or mexc explore.

Category	Module	Count	Description
Spot	marketv2	25	Market data, including public real-time streams; 21 do not require a key
Spot	kline	2	Klines; do not require a key
Spot	market-search	1	Market search
Spot	trade	12	Place, modify, cancel, and query orders; includes take-profit/stop-loss, trigger, trailing, and OCO orders
Spot	asset	32	Overview, balances, deposit/withdrawal and transfer history, inter-account transfers, bills, PnL
Spot	convert	8	Flash convert
Spot	dca	21	DCA / auto-invest
Spot	fee	9	Fee rate center
Spot	grid	16	Grid trading
Spot	account	1	Real-time stream of this account's trades
Contract	market	37	Public market data; do not require a key
Contract	trade	54	Orders, plan orders, stop-loss, trailing, leverage, margin, positions
Contract	account	26	Assets, fund records, multi-asset mode, fees, risk, preferences
Financial	flexible	18	Flexible products and auto-pledge
Financial	fixed	12	Fixed-term products and orders
Financial	contract-hold	10	Contract-hold yield
Financial	common	11	Common product, asset, position, and order queries
Activity	reward	4	Reward center
Activity	online	3	Activity discovery
Activity	participate	1	Activity registration
Futures market	market	3	Public real-time streams; do not require a key
Futures market	account	1	Real-time stream of this account's trades

If skills are not installed, you can still discover capabilities yourself: mexc explore browses everything, mexc contract market --help lists all capabilities in a module, and mexc schema outputs all capability IDs for scripting.

Option 2: Configure MCP

Configure by category, one entry per category:

{
  "mcpServers": {
    "mexc-spot": { "command": "mexc", "args": ["mcp", "spot", "--read-only"] }
  }
}

Most agents provide a configuration command, so you do not need to edit the file manually. Common agent setup examples:

claude   mcp add-json -s user mexc-spot '{"type":"stdio","command":"mexc","args":["mcp","spot","--read-only"]}'
codex    mcp add mexc-spot -- mexc mcp spot --read-only
gemini   mcp add -s user -t stdio mexc-spot mexc mcp spot --read-only
kiro-cli mcp add --name mexc-spot --command "mexc mcp spot --read-only" --scope global
code     --add-mcp '{"name":"mexc-spot","type":"stdio","command":"mexc","args":["mcp","spot","--read-only"]}'
cursor   --add-mcp '{"name":"mexc-spot","type":"stdio","command":"mexc","args":["mcp","spot","--read-only"]}'

Configure only what you need. Exposing all 307 capabilities to an agent lowers the chance it selects the right one and increases context usage.

args configuration	Coverage	Tool count
["mcp", "spot", "--read-only"]	All spot	123
["mcp", "contract", "--read-only"]	All contract	117
["mcp", "financial-openapi", "--read-only"]	Financial	51
["mcp", "activity", "--read-only"]	Activities and rewards	8
["mcp", "spot", "marketv2", "--read-only"]	Spot market only	22

For market and account queries only, the first entry is enough. File upload/download and real-time streams cannot be transported over MCP, so the tool count is lower than the capability count. All futures capabilities are real-time streams, so configuring MCP for them exposes no tools; use the terminal instead.

MCP Server Fails to Start

Replace mexc in the configuration with the full path to the binary. On Windows, a full path is required. On macOS and Linux, agents launched from a terminal usually do not need this; agents launched from the Dock or Launchpad may, especially Claude Desktop.

npm root -g     # Append /@mexc-ai/cli-<platform>/bin/mexc to the output directory

<platform> values include win32-x64, darwin-arm64, linux-x64, and others. On Windows, the filename includes .exe, and backslashes must be escaped in JSON:

"command": "C:\\Users\\YourUsername\\AppData\\Roaming\\npm\\node_modules\\@mexc-ai\\cli-win32-x64\\bin\\mexc.exe"