MCP: Providing standardized external tool access for large language models

MCP ( Model-Context Protocol ) is an innovative open protocol designed to provide secure and standardized access to external tools, data sources, and services for large language models. Unlike traditional plugin systems, MCP adopts a client-server architecture that enables seamless integration between models and external systems through a standardized protocol interface.

The Main Advantages of MC

1. Unified Interface: MCP defines standardized protocol specifications, avoiding the redundancy of independent integration solutions for each application development.

2. Security Assurance: Strict permission control and sandbox mechanisms ensure the secure access of external tools.

3. Highly scalable: Supports a variety of tool types, from simple API calls to complex data processing workflows.

4. Cross-platform compatibility: Any client that supports MCP can use the compatible service to achieve true interoperability.

The Integration of Web3 and MCP

In the Web3 domain, MCP can provide rich blockchain interaction capabilities for large language models:

• Asset Inquiry: View balances and transaction records of various tokens
• On-chain operations: sending transactions, deploying and calling smart contracts
• DeFi Integration: Interact with decentralized exchanges, lending protocols, liquidity mining, and other DeFi applications.
• Cross-chain operations: Support for multi-chain asset management and cross-chain transfers
• NFT Management: Query, Transfer, Trade Non-Fungible Token Assets

This article will build a basic Web3 MCP service using Node.js and TypeScript, delving into the working principles and best practices of MCP.

Build Web3 MCP Service

1. Project Initialization

First, create a project directory and initialize the npm project:

bash mkdir web3-mcp-demo cd web3-mcp-demo npm init -y

Install the necessary dependencies:

bash npm install @anthropic-ai/sdk ethers typescript @types/node npm install --save-dev ts-node

Configure TypeScript:

Create a tsconfig.json file and configure it according to the project requirements.

2. Write MCP server

Create a simple MCP tool, taking the getBalance function of the Sepolia testnet as an example:

typescript import { ethers } from 'ethers'; import { Server } from '@anthropic-ai/sdk/mcp';

// Initialize provider const provider = new ethers.providers.JsonRpcProvider('YOUR_RPC_URL');

// Create MCP tool const server = new Server();

server.tool({ name: 'getBalance', description: 'Check the ETH balance of a specified address on the Sepolia testnet', parameters: { type: 'object', properties: { address: { type: 'string', description: 'The Ethereum address to query' } }, required: ['address'] }, handler: async ({ address }) => { const balance = await provider.getBalance(address); return ethers.utils.formatEther(balance); } });

// Start the server server.listen(3000, '127.0.0.1', () => { console.log('MCP server has started, listening on port 3000'); });

// Error Handling process.on('uncaughtException', (error) => { console.error('unhandled exception:', error); });

process.on('unhandledRejection', (reason, promise) => { console.error( 'Unhandled Promise rejection:', reason); });

3. Service Debugging

Compile TypeScript code:

bash npx tsc

Debugging with MCP Inspector:

bash npx @anthropic-ai/sdk/mcp/inspector

4. Function Expansion

Further improve the service by adding more features:

• Supports multiple EVM-compatible networks
• Query gas fees
• Send Transaction
• Query token information and transaction records

5. Integrate in Cursor IDE

Cursor is a smart IDE that supports MCP integration, built on VSCode. By integrating the developed MCP services into Cursor, you can interact with the codebase using natural language.

Practical Application Examples

In the AI assistant of Cursor, you can easily perform the following operations:

1. Check address balance: "Check balance of address 0xE21E97Ad8B527acb90F0b148EfaFbA46625382cE on the Sepolia testnet"

2. Send Transaction: "Transfer 0.1 ETH to address 0x2c1d9ef7ccede70d77e6038701cd63138dd920a0"

Future Outlook

The combination of MCP and Web3 offers developers a broad range of application prospects. Possible future development directions include:

• Support for more blockchain networks (such as Bitcoin, Solana, Tron, etc.)
• Achieve cross-chain operations, such as cross-chain asset conversion.
• Integrate more DeFi protocols and NFT marketplace features

The integration of MCP and Web3 opens up a new world full of possibilities, where complex blockchain operations can be achieved through simple dialogues, greatly lowering the barrier to entry for Web3 applications.