home / mcp / akshare_mcp mcp server
Provides AKShare data interfaces as an MCP Server with configurable tool exposure and markdown output support.
Configuration
View docs{
"mcpServers": {
"wukan1986-akshare_mcp": {
"command": "D:\\\\Users\\\\Kan\\\\miniconda3\\\\envs\\\\py312\\\\python.exe",
"args": [
"-m",
"akshare_mcp",
"--format",
"markdown"
]
}
}
}You run an MCP Server that exposes AKShare data interfaces for easy access by MCP clients. This setup lets you enable or disable interfaces, format outputs, and manage how you route thousands of data endpoints through a single server.
Install and start the server, then connect your MCP client to the local runtime. You can enable only the interfaces you plan to use to manage tokens and performance. The server can run in a local Python environment and expose AKShare data endpoints as tools your MCP clients can request. To select which interfaces are available, you will edit the configuration and specify the interfaces you want to expose.
# Prerequisites: ensure Python is installed
# Create or activate a Python virtual environment
# Then install the MCP package
pip install akshare_mcp
# Verify installation and see help/usage
python -m akshare_mcp -hThe server is configured by running a local Python interpreter that launches the MCP process. A typical runtime command is shown below. This config runs the MCP server in markdown format by executing the module directly.
{
"mcpServers": {
"akshare_mcp": {
"command": "D:\\Users\\Kan\\miniconda3\\envs\\py312\\python.exe",
"args": [
"-m",
"akshare_mcp",
"--format",
"markdown"
]
}
}
}If you need to customize which AKShare interfaces are exposed, you can edit a Python config file to list the desired interfaces. The available interfaces can be referenced in AKShare’s data index. You can also run the server with a custom config path if you prefer a separate configuration file.
To start the server with a separate config file, you can pass a custom path as shown here.
python -m akshare_mcp --config D:\config.pyThere are 1000+ interfaces exposed as tools. Exposing all tools can consume tokens, so you should enable only the interfaces you need. Some MCP clients support a limited number of tools; plan accordingly (for example, Trae may support around 40 tools). Enable the exact tools you plan to use to optimize performance and token usage.