myquant-mcp

name	myquant-mcp
description	Calls this repository's MyQuant HTTP tools via bundled scripts/client.py (no direct MCP wiring); supports protected account, trading, and stock_do_t when the user passes --auth-token. Use when the user mentions MyQuant MCP, HTTP API, client.py, or skill helper scripts.

MyQuant MCP

Use the bundled scripts/client.py helper when the user wants skill-based access to this repository's MyQuant HTTP API.

What this skill does

• Query market data, fundamental data, account tools, trading tools, and the higher-level stock_do_t helpers from a running MyQuant server.
• List available tools directly from the live server.
• Explain how to pass --url when the server is not running on http://localhost:8001.
• Explain how to pass --auth-token explicitly for protected tools.
• Save API results to JSON or CSV through the helper script.

HTTP 工具覆盖（与 MCP 一致）

• 服务端注册的全部工具（当前共 61 个：行情 12、基本面 31、账户/成交 4、交易 11、stock_do_t 系 3）都走同一 HTTP 工具接口；任意工具名均可作为 client.py 的第一个位置参数调用。
• 权威清单：在仓库根目录执行 python skills/myquant-mcp/scripts/client.py --list-tools（需服务已启动），输出含 Public / Protected 分组；--info <tool_name> 查看该工具的参数 schema（与 MCP 相同）。
• scripts/client.py 里的 DATA_TOOLS：仅为行情 + 基本面共 43 个工具的离线说明，用于服务不可达时 --list-tools 的回退展示；不包含账户、下单、stock_do_t*——这些必须连上服务后用 --list-tools / --info 或查仓库 server/tools/{execution,trading,do_t}.py。
• 鉴权：凡带服务端 auth_token 参数的工具（账户、交易、stock_do_t* 共 18 个）在命令行用 --auth-token <MCP_AUTH_TOKEN>；行情与基本面工具无该参数。默认 REQUIRE_AUTH_TOKEN=true 时，未配置或错误的 MCP_AUTH_TOKEN 会导致受保护工具失败；若部署改为 REQUIRE_AUTH_TOKEN=false，以服务端实际校验逻辑为准。
• 调用方式：默认 POST JSON 到 /api/v1/tools/{tool_name}；需要 query 形式时用 --get（见 client.py --help）。

Important limits

• This skill uses scripts/client.py, which talks to the HTTP REST API.
• HTTP and MCP now expose the same tool set.
• Protected account and trading tools require the same auth_token as MCP.
• Because the skill may be installed into another agent workspace, it should not assume it can read this repo's .env.
• Pass --auth-token <MCP_AUTH_TOKEN> explicitly on the command line for protected tools.
• Trading APIs are more sensitive than pure data APIs, so keep token handling explicit and avoid embedding tokens in the skill itself.

Quick workflow

1. cd to the repository root (paths below assume skills/myquant-mcp/ lives under that root).
2. Confirm the server is running with python server.py.
3. Use python skills/myquant-mcp/scripts/client.py --list-tools to inspect tools from the live server.
4. Use python skills/myquant-mcp/scripts/client.py <tool_name> ... to call public tools.
5. Use python skills/myquant-mcp/scripts/client.py --auth-token <MCP_AUTH_TOKEN> <tool_name> ... for protected account, trading, and stock_do_t tools.
6. Add --url http://host:8001 if the server runs elsewhere.

Common examples

python skills/myquant-mcp/scripts/client.py --list-tools
python skills/myquant-mcp/scripts/client.py --info history
python skills/myquant-mcp/scripts/client.py history --symbol SHSE.600000 --frequency 1d --start-time 2024-01-01 --end-time 2024-12-31
python skills/myquant-mcp/scripts/client.py stk_get_daily_valuation --symbols SHSE.600519 --fields pe_ttm,pb_lyr
python skills/myquant-mcp/scripts/client.py --auth-token YOUR_TOKEN get_positions
python skills/myquant-mcp/scripts/client.py --auth-token YOUR_TOKEN stock_do_t --symbol SZSE.000001 --direction buy_then_sell --volume 1000 --expire-seconds 30 --entry-trigger-price 10.00 --entry-order-price 10.00 --take-profit-pct 3 --stop-loss-price 9.70

References

• Read the repository root README.md for direct MCP setup in Cursor, Claude Code, and OpenClaw.
• Read references/mcp-config.md for bundled config files and when to prefer the skill path instead.