QMT MCP Server

QMT MCP Server is an MCP server that interfaces with a locally running MiniQMT trading environment to bring stock trading functions to large language models. Built for quantitative developers and retail traders operating within Chinese equities markets, it bridges AI assistants with the XtTrader QMT system. Through this interface, models can read account balances, inspect current position details, place market or limit buy and sell orders, and submit order cancellations. Users interact with trading workflows using natural language prompts while retaining local oversight of account settings. The server requires a local MiniQMT installation alongside an active trading account, reading configuration parameters from a local YAML file. It enables automated execution loops and combined research-to-execution pipelines when paired alongside screening and analysis tools, giving AI agents operational agency over trading accounts under specified execution constraints.

Category: Developer Tools & Code Intelligence

Tags: algorithmic-trading, finance, qmt, trading

Visit QMT MCP Server

How to install and configure QMT MCP Server

1. Ensure Python 3.10 or higher is installed and the MiniQMT trading system is actively running locally. 2. Install the uv package manager. 3. Clone the repository: bash git clone https://github.com/nnquant/qmt-mcp-server 4. Install project dependencies: bash uv sync 5. Run the server to generate initial configuration prompts: bash uv run main.py Enter the MiniQMT path and funding account when prompted, which saves to xttrader.yaml. 6. Connect your MCP client (such as Cursor) using an SSE endpoint in your client configuration: json { "mcpServers": { "qmt-mcp-server": { "url": "http://localhost:8001/sse" } } }

What you can do with QMT MCP Server

• Checking portfolio holdings and available cash balances via natural language prompts in an MCP client. - Submitting limit and market orders for Chinese stocks using specific exchange suffixes like .SH or .SZ. - Cancelling active unfilled stock orders across open trading sessions without opening the desktop broker interface. - Calculating dynamic position sizing based on available account cash directly before placing buy transactions. - Combining research agents with trading execution to screen target equities and execute trades in a unified workflow.

Key facts

• https://github.com/nnquant/qmt-mcp-server
• Developer Tools & Code Intelligence, Finance, Crypto & Payments
• algorithmic-trading, finance, qmt, trading

Part of MCP Servers

What is QMT MCP Server?

QMT MCP Server is a Model Context Protocol server that connects AI clients to a locally running MiniQMT trading instance, enabling large language models to query accounts and execute stock orders.

What prerequisites are needed to use QMT MCP Server?

You need Python 3.10 or higher, the uv package manager, approved QMT trading permissions from your brokerage, and an active, running MiniQMT desktop application on the local system.

How should stock codes be specified when ordering?

Stock codes must include their respective exchange suffixes, such as 600000.SH or 000001.SZ. Relying purely on stock names may lead to incorrect ticker identification depending on the underlying AI model.

Which MCP clients work with QMT MCP Server?

Any MCP client supporting Server-Sent Events (SSE) network endpoints can connect to the server. The documentation demonstrates integration with Cursor using its standard mcpServers configuration file.