An open-source toolkit for connecting AI agents to EVM protocols. Now, any agent, using any model can autonomously perform various EVM actions:

• Check ERC20 token balances
• Transfer ERC20 tokens
• Launch tokens on Four.meme
• Get token pricing data
• Fetch trending tokens
• Monitor market movements
• Access DeFi protocols
• And more...

Anyone - whether an SF-based AI researcher or a crypto-native builder - can bring their AI agents trained with any model and seamlessly integrate with EVM-compatible blockchains.

🔧 Core Blockchain Features

• Token Operations

  • Check ERC20 token balances
  • Transfer ERC20 tokens
  • Launch new tokens on Four.meme (BSC)
  • Bridge tokens across chains

• Market Data Integration

  • CoinGecko Pro API integration
  • Real-time token price data
  • Trending tokens and pools
  • Top gainers analysis
  • Token information lookup

• DeFi Data Access

  • Protocol TVL tracking with DeFiLlama
  • Price feeds and aggregation
  • Token trading data

• Social Data

  • Elfa AI social media insights
  • Trending tokens by social activity
  • Smart mentions tracking
  • Social account analysis

🤖 AI Integration Features

• LangChain Integration

  • Ready-to-use LangChain tools for blockchain operations
  • DynamicStructuredTool-based implementation
  • Zod schema validation
  • Comprehensive error handling

• Vercel AI SDK Integration

  • Vercel AI SDK for AI agent integration
  • Framework agnostic support
  • Quick and easy toolkit setup

• MCP Integration

  • Support for the Model Context Protocol
  • Action-based architecture
  • Structured schema definitions

• AI Tools

  • DALL-E integration for image generation
  • Natural language processing for blockchain commands
  • Price feed integration for market analysis

📦 Installation

npm install evm-ai-kit

Quick Start

import { EvmAgentKit, createEvmTools } from "evm-ai-kit";

// Initialize with private key and optional RPC URL
const agent = new EvmAgentKit(
  "your-wallet-private-key",
  "https://rpc-url.example.com",
  { OPENAI_API_KEY: "your-openai-api-key" } // optional config
);

// Create LangChain tools
const tools = createEvmTools(agent);

Usage Examples

Check ERC20 Token Balance

const balance = await agent.getBalance("0x1234567890123456789012345678901234567890");
console.log("Token Balance:", balance);

Transfer ERC20 Tokens

const txHash = await agent.transfer(
  "0x1234567890123456789012345678901234567890", // to address
  1.5, // amount
  "0xabcdef1234567890abcdef1234567890abcdef12" // token address (optional for native token)
);
console.log("Transfer Transaction:", txHash);

Launch a Token on Four.meme (BSC)

const result = await agent.createFourMemeToken(
  FOUR_MEME_ADDRESSES.BSC, // Token Manager Address
  0.1, // Fee in BNB
  "My EVM Token", // name
  "MET", // short name/symbol
  "An awesome EVM token for testing", // description
  "https://example.com/token-image.png", // image URL
  Date.now() + 3600 * 1000, // launch time (1 hour from now)
  "Meme", // category
  "https://mytoken.com", // website (optional)
  "https://twitter.com/mytoken", // Twitter URL (optional)
  "https://t.me/mytoken" // Telegram URL (optional)
);

console.log("Token Creation Result:", result);

Get Token Price Data from CoinGecko

const priceData = await agent.getTokenPriceDataUsingCoingecko(
  "0x1234567890123456789012345678901234567890", // Token address
  "0xabcdef1234567890abcdef1234567890abcdef12" // Another token address
);
console.log("Token prices:", priceData);

Get Trending Tokens

const trendingTokens = await agent.getTrendingTokens();
console.log("Trending tokens:", trendingTokens);

Get Protocol TVL from DeFiLlama

const tvl = await agent.fetchProtocolTvl("uniswap");
console.log("Uniswap TVL:", tvl);

Creating a New Tool

Here's a complete example of creating a new tool in EVM Agent Kit:

1. Create the Tool

src/tools/my_tool/my_function.ts:

import { EvmAgentKit } from "../../agent";
import { Address } from "viem";

/**
 * Description of what your tool does
 * @param agent - EvmAgentKit instance
 * @param param1 - Description of first parameter
 * @param param2 - Description of second parameter
 * @returns Promise resolving to the result
 */
export async function my_function(
  agent: EvmAgentKit,
  param1: string,
  param2?: Address,
): Promise<any> {
  // Implementation of your tool
  // Use agent.connection, agent.wallet, etc. to interact with blockchain
  
  return result;
}

src/tools/my_tool/index.ts:

export * from "./my_function";

2. Create a LangChain Tool

src/langchain/my_tool/my_tool.ts:

import { Tool } from "langchain/tools";
import { EvmAgentKit } from "../../agent";

export class MyTool extends Tool {
  name = "my_tool_name";
  description = `Description of what your tool does.
  
  Inputs (JSON string):
    - param1: string, description (required)
    - param2: string, description (optional)`;

  constructor(private agent: EvmAgentKit) {
    super();
  }

  protected async _call(input: string): Promise<string> {
    try {
      const parsedInput = JSON.parse(input);
      const { param1, param2 } = parsedInput;
      
      if (!param1) {
        throw new Error("param1 is required.");
      }
      
      const result = await this.agent.myFunction(param1, param2);
      
      return JSON.stringify({ status: "success", data: result });
    } catch (error: any) {
      return JSON.stringify({
        status: "error",
        message: error.message,
      });
    }
  }
}

3. Create an Action

src/actions/my_tool/my_action.ts:

import { Action } from "../../types/action";
import { EvmAgentKit } from "../../agent";
import { z } from "zod";

export const myAction: Action = {
  name: "MY_ACTION",
  similes: ["my action", "perform my action", "do my action"],
  description: "Description of what your action does.",
  examples: [
    [
      {
        input: { param1: "value1", param2: "value2" },
        output: { 
          status: "success", 
          data: { result: "example result" },
          message: "Action completed successfully" 
        },
        explanation: "Explanation of what this example demonstrates.",
      },
    ],
  ],
  schema: z.object({
    param1: z.string().describe("Description of param1"),
    param2: z.string().optional().describe("Description of param2"),
  }),
  handler: async (agent: EvmAgentKit, input: Record<string, any>) => {
    const { param1, param2 } = input;
    
    const result = await agent.myFunction(param1, param2);
    
    return {
      status: "success",
      data: result,
      message: "Action completed successfully",
    };
  },
};

4. Add to Agent

src/agent/index.ts:

import { my_function } from "../tools";

export class EvmAgentKit {
  // Existing properties and methods...
  
  async myFunction(param1: string, param2?: Address): Promise<any> {
    return my_function(this, param1, param2);
  }
}

Supported Chains

The toolkit supports all EVM-compatible chains. Specific features like Four.meme token creation are currently only available on specific chains (BSC for Four.meme).

Dependencies

The toolkit relies on several key EVM libraries:

• viem for blockchain interactions
• @langchain/core for AI agent tools
• zod for schema validation
• CoinGecko API for market data

Contributing

Contributions are welcome! Please feel free to submit a Pull Request. Refer to CONTRIBUTING.md for detailed guidelines on how to contribute to this project.

License

Apache-2 License

Security

This toolkit handles private keys and transactions. Always ensure you're using it in a secure environment and never share your private keys.