Provides MCP access to Wind API data and date calculations for LLM-driven workflows.
Configuration
View docs{
"mcpServers": {
"abuttoncc-wind-mcp": {
"url": "http://localhost:<port>/mcp/"
}
}
}Wind MCP Server exposes a standard MCP interface that lets large language models directly call Wind API functions. You install the server, expose it on your network, and integrate it with any MCP-compatible client so the model can query real-time Wind data and perform financial calculations as if it were a dedicated analyst tool.
You connect an MCP client to the Wind MCP Server and start issuing tool calls that map to WindPy functions. The server exposes both data retrieval and date-related helpers, so your model can fetch daily time series data, cross-sectional values for specific dates, and calendar-based date calculations. Start by configuring a client to reach the server and then invoke the available wind_mcp tools just like you would call any other MCP endpoint.
Key usage patterns you can perform after setup: retrieve historical price data for multiple assets over a date range, obtain snapshots of market metrics for several assets on a single date, and compute trading day offsets or counts within a given interval. The tools are designed to be straightforward function calls that your model can compose into complex queries and analyses.
Prerequisites: ensure you are running on Windows with Wind金融终端 installed and able to log in. You will also need Python installed on the machine that will run the MCP server.
Step 1: Install Python dependencies.
pip install -r requirements.txtStep 2: Start the MCP server so it is reachable on your LAN.
python src/wind_mcp_direct_server.py --host 0.0.0.0 --port 8888
```
* `--host 0.0.0.0` allows other devices on the LAN to access the server.
* `--port 8888` can be changed to your preferred port.Server connection options: you can expose the MCP server via HTTP or run it locally as a stdio process. Use the following configuration to connect from an MCP client.
{
"mcpServers": {
"wind_mcp": {
"url": "http://localhost:8888/mcp/",
"transport": "streamable-http",
"type": "http",
"args": []
}
}
}If you prefer running the server as a local process (stdio) with Python, use this configuration snippet.
{
"mcpServers": {
"wind_mcp": {
"type": "stdio",
"name": "wind_mcp",
"command": "python",
"args": ["src/wind_mcp_direct_server.py", "--host", "0.0.0.0", "--port", "8888"]
}
}
}Note: start commands shown above map to the examples in the setup steps. The HTTP URL uses the port you choose when starting the server. If you need to adjust the host, port, or transport, modify the corresponding fields in your client configuration.
Security and maintenance: keep the Wind terminal connected and logged in, ensure the server port is firewalled from unauthorized access on public networks, and monitor logs in the logs directory to catch any runtime issues early.
This server is designed to be lightweight and easy to integrate with MCP-compatible clients. All tool functions are exposed as standard MCP endpoints, enabling you to compose complex queries that leverage Wind data alongside your model’s reasoning capabilities.
If the server does not respond, verify that Wind terminal is logged in, the Python environment is active, and the chosen port is not blocked by a firewall. Check the logs/ directory for errors and ensure the host value matches the address used by your client.
Fetches the server's current date, with optional formatting.
Search the WindPy documentation for relevant API usage hints.
Retrieve daily time-series data for multiple instruments over a date range.
Retrieve daily snapshot data (cross-sectional) for multiple instruments on a specific date.
Fetch sequence data for constituents of a market index or sector over time.
Get trading days within a specified date range.
Calculate a future or past trading date offset from a base date.
Count the number of trading days in a given date interval.