What is drand-mcp-server?

Drand-mcp-server is a server that integrates with the Model Context Protocol (MCP) to fetch verifiable random numbers from the drand network. It allows AI applications to use secure, verifiable randomness as an input seed for model-driven flows.

How to use drand-mcp-server?

1. Installation: The server can be run using npx or by building from source locally.
   • VS Code Integration: Configure the server in .vscode/mcp.json with the npx command.
   • Claude Desktop Integration: Add the server to Claude's config.
2. Usage: Use the provided tools (e.g., get-randomness-latest, get-randomness-by-time, get-randomness-by-round) to fetch random numbers based on different criteria.
3. Building from Source: Clone the repo, install dependencies, build, and run the application.

Key Features of drand-mcp-server?

• Integration with drand network to fetch verifiable random numbers.
• Supports MCP for seamless integration with AI applications.
• Tools to fetch randomness by time, round, or the latest available beacon.
• Configurable for use with VS Code and Claude Desktop.

Use Cases of drand-mcp-server?

1. Repeatable, random sampling of input data for AI models.
2. Verifiable interaction with other MCP servers (e.g., for reward distribution).
3. Validating the output of random processes using historical drand beacons.

FAQ from drand-mcp-server?

• Is randomness truly verifiable?

  Yes, drand uses distributed randomness generation across multiple parties to ensure verifiability.

• Can I select items from a list with this server?

  A roadmap item mentions this feature, but it is not yet implemented.

• What versions of Node.js are supported?

  Node.js v21 or later is required to use the fetch API.

drand-mcp-server 🎲

Use verifiable randomness in your AI application. This Model Context Protocol (MCP) server enables you to get a random value from the drand network, verify its validity and use it as an input seed to your model-driven flows!

Use Cases

• repeatable, random sampling of input data
• interaction with other MCP servers in a verifiable manner (e.g. paying out rewards based on a prompt)
• verifying the output of another random process using historical drand beacons

Prerequisites

• a relatively recent version of node (v21+ - fetch is required)

Installation

You can run the MCP server either using npx or after building locally.

Usage with VS Code

Create a file called .vscode/mcp.json in your workspace (or in your home directory) and add the following code:

{
  "servers": {
    "drand": {
      "command": "npx",
      "args": [
        "drand-mcp-server"
      ]
    }
  }
}

For additional info, see the VS Code docs on MCP

Usage with Claude

You can run the drand-mcp-server alongside claude desktop by adding the following to your config:

{
  "mcpServers": {
    "drand": {
      "command": "npx",
      "args": [
        "drand-mcp-server"
      ]
    }
  }
}

Tools

The following tools are available from the MCP server

Name	Params	Description
get-randomness-latest	none	fetches the latest available beacon from drand quicknet
get-randomness-by-time	time in milliseconds	fetches the randomness beacon emitted at or just before the time
provided
get-randomness-by-round	round	fetches the randomness beacon emitted with a given round number

Building from source

• install dependencies with npm install
• build the application with npm run build
• run the application with either npm start or node ./dist/index.mjs

You can also configure VS Code and Claude as above, replacing the command/args with the following:

  "command": "node",
  "args": ["/path/to/my/project/drand-mcp-server/dist/index.mjs"]

Roadmap

• [x] fetch latest randomness
• [x] fetch randomness by round
• [x] fetch randomness by time
• [ ] select items from a list