Wmshare
Python SDK MCP server: offline catalog + 2,752 precomputed A-share/HK/US metrics for AI agents.
At a glance
Not tested yet- GitHub stars
- 24
- Weekly downloads
- 2,808
- Last commit
- 2026-11-19
- Licence
- BSD-3-Clause
- Transport
- stdio
- Auth
- none
uvx wmshare
What it does
Wmshare is a Python SDK for China A-share, Hong Kong and US market data. It covers k-line, intraday, money flow, statements, a screener and whole-market batch requests. It also ships a local MCP server so an AI agent can call its data interfaces. It is aimed at Python quant developers and at people who want market data inside an AI assistant.

Tools it gives your agent
The sources describe a local MCP server with 20 task tools. They do not list the individual tool names, so the tools cannot be named here. The sources state that all 155 friendly APIs, 138 registered plus 17 composite interfaces, can be reached through wmshare_call(target=..., mode="api"), and that the macro domain is available only by that path.
Quick start
- Install the SDK with MCP support.
- Run the setup command and paste the API Key when prompted.
- Run the doctor check.
pip install "wmshare[mcp]" wmshare-mcp setup --client-id <AppId> --client-secret-stdin --market cn wmshare-mcp doctor
Add it to Claude Code from a terminal:
claude mcp add wmshare -- uvx wmshare
Or add this to .mcp.json in your project root (Claude Code) or .cursor/mcp.json (Cursor):
{
"mcpServers": {
"wmshare": {
"command": "uvx",
"args": [
"wmshare"
]
}
}
}Environment variables
WMSHARE_CLIENT_ID: the AppId (Client ID) value from the open platform.WMSHARE_CLIENT_SECRET: the API Key (Client Secret) value from the open platform.
Example prompts
- Show me the TTM dividend yield for 600519.
- Which stocks have a three-year average dividend yield above 3% and ROE above 15%?
- What is the PE percentile over the last ten years for this stock?
- How did the whole market move today?
What to watch out for
- Credentials are required on open.winmale.com. The AppId and API Key are both needed.
- The offline demo client uses a frozen snapshot and is not live data. Its screener query needs real credentials.
- The valuation series used for backtests is at most 3 years.
- Server-side precomputed metrics are billed per symbol when you evaluate them yourself, so check the catalog before writing code.
- The sources state no other limits.
Not tested by tools4agents yet. Read the source repository before you install it.
Similar tools
Everything
The official test server. Exercises every part of the protocol, so it is the fastest way to check your setup works at all.
Filesystem
Gives an agent scoped read and write access to directories you name explicitly. The usual first MCP server people install.
Git
Read, search, and manipulate local Git repositories. Lets an agent inspect history and diffs without shelling out.
Sequential Thinking
Gives a model a scratchpad for multi step reasoning, letting it revise earlier steps instead of committing to a first answer.