Maverick

Overview

Maverick is a platform to create AI agents that can autonomously trade memecoins for you based on your selected preferences and insights from top traders. Built to simplify meme coin trading, the agent leverages real-time on-chain and social data for intelligent, automated decision-making.

Website: get-maverick.vercel.app
API Docs: maverick-backend.onrender.com/api

---

Problem

Memecoin trading is highly volatile and hype-driven, often influenced by social media trends and Key Opinion Leaders (KOLs). Traders lack efficient tools to keep up with these rapid changes, missing out on opportunities.

---

Solution

Maverick allowes you to set up AI agents that automate the trading process by:

• Using AI to analyze market sentiment and trends.
• Following strategies of trusted traders via their blockchain addresses or Farcaster handles.
• Executing trades for you based on pre-set persona.

---

Tech Stack

• Frontend:
  • TypeScript
  • Next.js
• Backend:
  • TypeScript
  • NestJS
  • PostgreSQL
  • PrismaORM
  • Redis
  • Docker
  • BullMQ
• AI:
  • Google Gemini for trend prediction and decision-making.
• Blockchain:
  • Currently, agents are deployed on Solana.
  • Agent: Leveraging the 0xDisciple agent from FereAI for autonomous trading.

---

Current Progress

Backend

1. JWT-based authentication using Solana signature (Ed25519) verification.
2. Create an agent with selectable personas:
   • MOON_CHASER: Focuses on short-term gains from meme coins with high volatility.
   • MEME_LORD: Prioritizes meme popularity on Farcaster.
   • WHALE_WATCHER: Tracks and mimics strategies of large traders.
3. Fetch current holdings of an agent.
4. Fetch portfolio information of an agent.
5. Generate and dispatch trading signals (buy/sell/hold) for coins based on social (Farcaster) sentiment.
6. Integrated pipeline framework for coin signal processing with BullMQ.
7. Added endpoints to fetch processed coin signals.

Frontend

1. Improved landing page UI.
2. Completed onboarding flow for agent creation and setup.
3. Dashboard displays portfolio and trading signal data.
4. Periodic updates for real-time portfolio, holdings insights.

---

TODOs

• Fix create agent call and formatting of decision_prompt_pool and decision_prompt_portfolio.
• Migrate backend to NestJS and PostgreSQL.
  • Real-time agent holdings tracker via CRON job.
  • Analyze Farcaster posts:
    • Identify posts about coins in the agent's holdings.
    • Use AI to determine if the sentiment is positive, negative, or neutral.
    • Generate buy/sell/hold signals for agents holding those coins.
  • Add get portfolio API.
  • Add JWT-based user authentication.
  • Integrate Redis for signal management.
  • Add BullMQ-based pipeline for signal processing.
• Frontend integration and improvements:
  • Improve landing page UI.
  • Implement user onboarding.
  • Create agent.
  • Get holdings.
  • Get portfolio.
  • Get processed coin signals.
• Persona selection should influence how the agent weighs social signals. For example, MEME_LORD gives more emphasis on meme popularity.
• Implement refresh access token mechanism for authentication.
• KOL signal integration:
  1. Fetch and analyze posts from user-supplied KOL handles.
  2. Generate trading signals based on KOL discussions for both portfolio and potential coins.
• Integrate the BUY API when it's available from FereAI.
• [LONG TERM] Generalize the agent for all types of trading including crypto and stocks, beyond meme coins.

---

Stay tuned for updates as the project evolves! Contributions and feedback are welcome.