Mantle AI Agent SDK & Chat Interface

A production-ready infrastructure toolkit for building verifiable AI agents on the Mantle blockchain. This project provides a modular SDK for agent orchestration and a conversational frontend that bridges natural language to on-chain execution.

PS: AI Voice present in the Demo Video - Link

One Pager Pitch Link - Click Here

Project Overview

The Mantle AI Agent SDK allows developers to build agents that observe on-chain state, reason off-chain using Large Language Models, and execute verifiable actions on Mantle. By treating Mantle as the execution and audit layer, the SDK ensures that every AI decision is linked to an on-chain receipt.

Mantle AI SDK Npm Package

npm Package

Demo

Watch the AI Agent Chat Interface in action: Mantle AI Agent Demo

Key Features

Observe-Decide-Act Lifecycle: A structured orchestration flow that manages the full lifecycle of an autonomous agent.
Verifiable AI Reasoning: Every action taken by the agent is cryptographically linked to an off-chain Reasoning Hash stored on-chain for full auditability.
Automatic Asset Creation: On-demand deployment of ERC20 Tokens and ERC721 NFT Collections (with metadata/images) through natural language commands.
Batched Execution: Execute multiple complex on-chain actions in a single atomic transaction to save gas and ensure state consistency.
Pluggable LLM Providers: Built-in support for Gemini 1.5 Flash, with a swappable interface for other models like GPT-4 or Claude.
Mantle-Native Infrastructure: Fully optimized for Mantle's L1 Data Fee and high-throughput environment.

Deployed Infrastructure (Mantle Sepolia)

The following core infrastructure contracts are live and verified on the Mantle Sepolia Testnet:

Contract	Address
MantleAgentExecutor	`0x46Ac132785eC02B6d8903985eb934EbCED8b6ea3`
MantleAssetFactory	`0x6346c84E65db504faA791d87378D59d9E698743C`

Tech Stack

Blockchain: Mantle Network (High-throughput, low-fee L2 execution and audit layer).
AI Reasoning: Google Gemini 1.5 Flash (Low-latency, cost-effective reasoning engine).
Smart Contracts: Solidity (0.8.19), OpenZeppelin (Secure standard assets).
Frontend: Next.js 15, Tailwind CSS, Ethers.js v6.
Tooling: Hardhat (Deployment & Testing), TypeScript (Type-safe SDK).

How to Use the SDK

The SDK is designed to be integrated into any TypeScript project.

1. Initialize the Agent

import { Agent, GeminiProvider, MyCustomObserver } from "@mantle-agent-sdk/core";

const config = {
  agentId: "my-agent-001",
  mantleRpcUrl: "https://rpc.sepolia.mantle.xyz",
  executorAddress: "0x...",
  privateKey: "0x..."
};

const agent = new Agent(
  config,
  new GeminiProvider(process.env.GEMINI_API_KEY),
  new MyCustomObserver(config.mantleRpcUrl)
);

2. Run an Iteration

const result = await agent.runIteration("Observe the floor price and buy if it drops below 1 MNT.");
console.log("Transaction Hash:", result.transactionHash);

3. Create Assets Programmatically

import { AssetUtils } from "@mantle-agent-sdk/core";

// Encode an action to create a tokenconst action = AssetUtils.createTokenAction(
  factoryAddress,
  "My AI Token",
  "AIT",
  ethers.parseEther("1000000"),
  userAddress
);

Future Scope

Automatic Contract Creation & Management

The current SDK allows for the immediate deployment of standard ERC20 and ERC721 contracts. The future roadmap includes:

Dynamic Logic Synthesis: Agents that can write, compile, and deploy custom smart contract logic on Mantle based on evolving user requirements.
Autonomous Treasury Management: Agents that manage their own liquidity pools and rebalance assets across Mantle DeFi protocols.
Verifiable Reasoning Logs: Integration with decentralized storage (like IPFS or Mantle DA) to store full LLM reasoning chains, linked cryptographically to the reasoningHash emitted on Mantle.
Images