参见 https://www.quantconnect.com/docs/v2/ai-assistance/mcp-server/key-concepts

QuantConnect MCP 服务器

QuantConnect MCP 服务器是 AI（如 Claude 和 OpenAI o3 Pro）与我们的云平台交互的桥梁。配备我们的 MCP 后，AI 可以通过我们的 API 代表你执行任务，如更新项目、编写策略、回测以及将策略部署到生产线实时交易。

这是QuantConnect团队维护的官方MCP实现。我们建议使用官方版本以确保代码和API令牌的安全。我们的实现经过测试并通过docker化，便于跨平台部署。

开始

要将本地MCP客户端（如Claude Desktop）连接到QC MCP服务器，请按以下步骤操作：

1. 安装并打开Docker Desktop。
2. 安装并打开 Claude Desktop。
3. 在 Claude Desktop 中，点击 文件>设置>开发者 >编辑配置。
4. 编辑 claude_desktop_config.json 文件以包含以下 quantconnect 配置：

要获取你的用户ID和API令牌，请参见请求API令牌。

我们的MCP服务器支持多平台支持。

QuantConnect Remote MCP Server

The QuantConnect Remote MCP Server is the preferred way to connect AI assistants such as Claude to the QuantConnect platform.

There is nothing to install, build, or run. Point your MCP client at the hosted endpoint, authorize it against your QuantConnect organization, and the tools are available immediately.

• Documentation: https://www.quantconnect.com/docs/v2/ai-assistance/mcp-server/key-concepts
• Remote MCP endpoint: https://www.quantconnect.com/api/v2/mcp

Note: The remote MCP server requires a paid agent node (A1-1, A4-8, or A16-32) in your organization. The free A-MICRO node cannot connect. See Agent Nodes and Add Nodes.

Quick Start with Claude

The remote server speaks the standard MCP HTTP transport, so any MCP-capable Claude surface can use it. Pick the one you use.

Claude Desktop and claude.ai

1. Open Settings > Connectors (on the desktop app, File > Settings > Connectors).
2. Click Add custom connector, give it a name such as QuantConnect, and enter the server URL:

   https://www.quantconnect.com/api/v2/mcp
   

3. Click Connect. Your default browser opens.
4. Click Continue connecting, select your QuantConnect organization in the Authorize MCP Server form, and click Authorize.
5. Start a new chat. The QuantConnect tools are listed under the connector and can be enabled per conversation.

Claude Code (terminal)

Register the server once from any terminal:

claude mcp add --transport http quantconnect https://www.quantconnect.com/api/v2/mcp

Then start Claude Code, run /mcp, select quantconnect, and follow the browser authorization flow (Continue connecting > Authorize). Running /mcp again should show the server as connected with its tools listed.

Claude Code in Local Platform (VS Code extension)

1. Install the Claude Code extension in VS Code.
2. Create or open a QuantConnect project in Local Platform.
3. Open the Claude Code panel from the top navigation bar.
4. Enter /mcp to confirm the QuantConnect MCP Server is connected.

Try it

Once connected, send a simple prompt to verify the tools are working:

Read the open project.

Claude should call the read_open_project tool and describe the project. From there you can ask it to create projects, write and edit strategy files, run backtests, optimize parameters, or deploy to paper trading, all in plain language. For example:

Create a new Python project that buys SPY on the first trading day of each month,
backtest it from 2020 to today, and summarize the results.