project screenshot 1
project screenshot 2
project screenshot 3
project screenshot 4

Based Agents

Build & trade AI agents with unique personalities! Create autonomous traders that learn from Farcaster influencers like @vitalik. As a Commander NFT holder, you control your agent's on-chain decisions. Conservative or degen - you choose the strategy!

Based Agents

Created At

ETHGlobal Bangkok

Winner of

XMTP - Building Agents with MessageKit

Coinbase Developer Platform - CDP SDK Pool Prize

Prize Pool

Base - Best Use of OnchainKit 1st place

Project Description

AI Agent Builder & Marketplace: A Comprehensive Overview

What is it?

A decentralized platform that enables users to create, customize, and trade AI-powered autonomous trading agents. Each agent exists as a unique NFT (called a Commander NFT) and can execute on-chain trading strategies while learning from social media personalities.

Core Components

  1. Agent Creation System
  • Customizable Risk Profiles

    • Users can select from multiple risk levels (conservative to "degen")
    • Each profile comes with pre-configured trading parameters
    • Custom risk settings available for advanced users
  • Personality Engineering

    • Agents can be trained on specific Farcaster personalities
    • Learning algorithms analyze posting patterns, trading behavior, and market sentiment
    • Custom prompts allow fine-tuning of agent behavior and decision-making
  • Safety Mechanisms

    • Built-in stop-loss functionality
    • Automated position closing at target prices
    • Risk management protocols to prevent catastrophic losses
  1. Interactive Features
  • Natural Language Interface

    • Direct communication with agents through Converse App
    • Real-time trading suggestions and market analysis
    • Query historical decisions and performance metrics
  • Commander NFT System

    • ERC-721 tokens represent ownership of agents
    • Only token holders can modify agent parameters
    • Tradeable on secondary markets
    • Governance rights over agent behavior

Use Cases

  1. Passive Trading

    • Set up agents to manage portfolios automatically
    • Implement complex trading strategies without active management
  2. Strategy Development

    • Test and refine trading approaches
    • Learn from successful traders through agent personality modeling
  3. Asset Management

    • Delegate trading decisions to AI agents
    • Maintain control through Commander NFT ownership
  4. Trading Strategy Marketplace

    • Buy and sell successful agent configurations
    • Trade Commander NFTs for profit

Future Potential

  • Integration with additional blockchain networks
  • Enhanced AI models and learning capabilities
  • Expanded social media personality analysis
  • Advanced governance features
  • Cross-chain trading capabilities
  • Community-driven strategy development

How it's Made

Core Architecture

  1. Frontend
  • Built with React + Vite
  • Deployed to Vercel
  • Features a dashboard for agent creation, management, and listing
  • Makes use of Base Onchain components
  1. Backend Infrastructure The system relies on four main components:
  • XMTP Message-kit for secure agent-commander communication
  • Base Onchain AgentKit for executing on-chain actions
  • Hyperbolic Models for decentralized AI processing
  • Data Service for analyzing Farcaster & social media personalities
  1. Smart Contracts
  • Implements Commander NFTs (ERC-721)
  • Handles agent ownership and trading rights

Technical Requirements

  • Node.js 16+
  • Python 3.8+
  • Docker

Key Features Implementation

Agent Creation

  • Risk strategy selection (conservative to degen)
  • Personality engineering through custom prompts
  • Integration with Farcaster personalities
  • Safety controls implementation (exit targets and stop losses)

Agent Interaction

  • Integration with Converse App
  • Natural language processing for agent communication
  • Real-time trading features
  • Access control through Commander NFT
background image mobile

Join the mailing list

Get the latest news and updates