Mindcore MCP Server

MCP (Model Context Protocol) server for onchain chat, DeFi operations, and AI model integration.

Overview

This library provides a framework for AI-powered chat and DeFi operations (swap, transfer) via MCP-compliant tools. It is designed for seamless integration with AI models and LLM tool providers.

Related MCP server: agenti

Features

• AI Chat Agents: Analyst and Trader Copilot powered by OpenAI.

• DeFi Operations: Swap tokens (Uniswap), transfer ETH/ERC20.

• Intent Parsing: Natural language understanding for DeFi actions.

• MCP Tool Manifest: Ready for LLM/AI tool integration.

• Framework Agnostic: Use with Express, Fastify, h3, or any Node.js backend.

Installation

npm install mindcore-mcp

Usage

Basic Example

import { createChatHandler, swapHandler, transferHandler } from 'mindcore-mcp';

// Example: Express.js
app.post('/api/chat', createChatHandler());

// Example: Direct function call
const chatResponse = await chatHandler({
  messages: [{ role: 'user', content: 'Swap 1 ETH to USDC' }],
  agentId: 2
});

const swapResponse = await swapHandler({
  from: 'ETH',
  to: 'USDC',
  amount: '1',
  userAddress: '0xYourAddress'
});

Running the Server

You can run the Mindcore MCP Server as a standalone HTTP server:

npx mindcore-mcp

This will start the server on the default port (3000). You can set a custom port with the PORT environment variable.

MCP Tools

Environment Variables

Copy .env.example to .env and fill in your values.

Usage with LLM Tools

The Mindcore MCP Server exposes the following MCP-compliant tools for LLMs:

chat

Endpoint: POST /api/chat Input:

{
  "messages": [
    { "role": "user", "content": "Swap 1 ETH to USDC" }
  ],
  "agentId": 2
}

Output:

{
  "reply": "swap_intent",
  "swap": { "amount": "1", "from": "ETH", "to": "USDC" },
  "message": "You want to swap 1 ETH to USDC. Please confirm the swap."
}

swap

Endpoint: POST /api/swap Input:

{
  "from": "ETH",
  "to": "USDC",
  "amount": "1",
  "userAddress": "0xYourAddress"
}

Output: Returns transaction data for the swap.

transfer

Endpoint: POST /api/transfer Input:

{
  "token": "USDC",
  "amount": "10",
  "recipient": "0xRecipientAddress",
  "userAddress": "0xYourAddress"
}

Output: Returns transaction data for the transfer.

Registering Tools with LLMs

To use these tools with an LLM, add them to your tool manifest or configuration. Example (pseudo-config):

{
  "mcpServers": {
    "mindcore": {
      "command": "npx",
      "args": [
        "mindcore-mcp"
      ],
      "env": {
        "OPENAI_API_KEY": "your_openai_api_key",
        "RPC_URL": "your_rpc_url",
        "UNISWAP_ROUTER_ADDRESS": "your_router_address"
      }
    }
  }
}

Prompting Tips:

• Instruct the LLM to use the chat, swap, and transfer tools for onchain actions.

• Provide the tool descriptions and example calls in your LLM's tool registry.

Development

# Build the package
npm run build

# Run tests (if implemented)
npm test

License

MIT