AGI Solver

AGI Solver is a backend system written in TypeScript designed to reliably fulfill onchain interaction requests (Agent Generated Intents) for AI agents. It serves as a critical bridge between AI agents and blockchain interactions, ensuring reliable execution of complex onchain operations.

Overview

The AGI Solver system consists of two main components:

1. TypeScript Backend: Handles the complex logic of executing onchain interactions reliably, managing:

   • Transaction execution and retry logic

2. Smart Contract System (mock13/): Provides the onchain infrastructure for:

   • Publishing agent intents
   • Managing agent assets
   • Serving as a smart contract wallet for agents

Smart Contract Architecture

The contracts/ folder contains the onchain component of the AGI Solver system:

Key Contracts

• Mock13.sol: The main contract serving as a smart contract wallet for agents
• TokenA.sol & TokenB.sol: Test tokens for demonstrating trading functionality

Key Features

1. Intent Publication

   function publishAGI(
       uint8 intentType,
       address assetToSell,
       uint256 amountToSell,
       address assetToBuy
   )

   Agents can publish their trading intents through this function.

2. Asset Management

   • The contract acts as a secure wallet for agent assets
   • Supports ERC20 token interactions
   • Manages trade execution and asset transfers

Deployed Contracts (Base Sepolia)

• Mock13: 0x538Dd1dB653bbF7376CF8C57C6bF68805Cf01166
• TokenA: 0x011228A36559f2029982bB75947BD3CAc2Eb9fF9
• TokenB: 0x0B44519951121D60b3241272ADeBA7a944B63761

Integration with Semantic Layer

AGI Solver is designed to be integrated into the Semantic Layer ecosystem and will be a key component in the upcoming Silicon Valley Fun product release. This integration allows:

1. AI agents to focus on high-level decision making and reasoning
2. AGI Solver to handle all complex onchain interactions
3. Reliable execution of agent-generated intents
4. Seamless interaction with DEXes and lending protocols

Purpose

The project addresses several key challenges in onchain AI agent interactions:

• Complex transaction management
• Failed transaction handling
• Optimal execution strategies
• Gas optimization
• RPC reliability

By separating the execution layer (AGI Solver) from the reasoning layer (AI agents), we create a more robust and reliable system for onchain AI interactions.

Development

Prerequisites

• pnpm
• Foundry
• TypeScript

Setup

1. Clone the repository

2. Set up environment variables:

   cp .env.example .env

   cd contracts
   cp .env.example .env

3. Deploy contracts under root directory

   make deploy
   

4. Start solver

   make start
   

5. Publish an agi

   #e.g.
   make sellTokenA
   

Contributing

We welcome contributions! Please see our contributing guidelines for more details.

License

MIT