drand-mcp-server

drand-mcp-server is a Model Context Protocol (MCP) server designed to provide verifiable randomness for AI applications. It fetches random values from the drand network and supports verifying their validity, so it can be used as an input seed for model-driven processes.

Purpose

This server is suitable for scenarios that require introducing verifiable randomness into AI applications, such as:

• Reproducible random sampling of input data;
• Interacting with other MCP servers in a verifiable way (for example, issuing rewards based on prompts);
• Using historical drand beacons to verify the output of another random process.

Main capabilities

Through this MCP server, the following tools can be used:

• Fetch the latest random number;
• Fetch a random number by round;
• Fetch a random number by time;
• Select an item from a list.

Usage

Running this MCP server requires a relatively recent Node.js version (v21+, because fetch is required). It can be run via npx, or built locally and then run.

Using with VS Code

In the workspace (or home directory), create .vscode/mcp.json file, and add the corresponding configuration. For more information, refer to the VS Code documentation on MCP.

Using with Claude

The corresponding content can be added to the Claude Desktop configuration so that drand-mcp-server can be run together with Claude Desktop.

Building from source

1. Use npm install to install dependencies;
2. Use npm run build to build the application;
3. Use npm start or node ./dist/index.mjs to run the application. VS Code and Claude can also be configured as described above, and the command/arguments can be replaced with the run command after the above build.

Applicable scenarios

• AI applications need verifiable random number input;
• Verifiable interaction with other MCP servers is needed;
• The correctness of the output of another random process needs to be verified.

Limitations

• Requires Node.js v21 or later;
• Specific configuration details (such as the complete configuration code for VS Code and Claude) are not fully provided in the materials, and the corresponding documentation should be consulted;
• Information such as other features, fees, pricing, platforms, effects, or usage steps is not yet provided.