A developer going by "geeleeto" has published full source code for Dexterfun, a complete Solana meme-coin sniper designed to detect new Pump.fun launches and execute trades before the crowd arrives. The project dropped on DEV.to August 15th with a Node.js backend, React dashboard, and everything you need to run your own launch-detection pipeline.

Architecture Overview

The stack is refreshingly conventional for crypto tooling: Node.js powers the backend with Express handling REST endpoints and Socket.io pushing real-time updates. On the frontend, React pairs with Vite for fast development iteration. Data persistence uses SQLite for structured state while Redis handles caching and hot-path data that needs sub-millisecond access times. The system monitors the Pump.fun ecosystem continuously, scanning for newly created tokens the moment they hit the platform. When a launch is detected, the bot doesn't just fire blindly—it runs new coins through a multi-stage rug-filter pipeline designed to weed out honeypots, liquidity traps, and obvious scam setups before any capital gets committed.

Execution Via Jito Bundles

Trade execution happens through Jito bundles, which batch multiple transactions together for atomic execution. This matters because in meme-coin land, being first is everything. A bundle that lands your buy order simultaneously with (or just after) the liquidity provision creates a much better entry price than chasing a coin that's already moved 200% on fresh volume. The exit strategy uses a phase-based state machine to manage positions post-entry. Rather than hard-coding take-profit targets, the bot transitions through defined phases—early position building, monitoring, partial exits at resistance levels, and final liquidation—based on observable market conditions rather than simple time or price gates.

Backtesting Included

One thing that separates this from weekend-project bots is the inclusion of backtest tooling. You can replay historical Pump.fun launches against your filter parameters to see which configurations would have saved you from rugs versus which ones would have filtered out legitimate opportunities. This kind of systematic testing is what separates traders who lose money from those who document why they lost it.

Key Takeaways

  • Full-stack implementation with Node.js, React/Vite, Express/Socket.io, SQLite, and Redis
  • Multi-stage rug-filter pipeline for pre-trade risk assessment on new Pump.fun launches
  • Jito bundle execution for competitive trade timing against other participants
  • Phase-based state machine architecture for systematic exit management rather than emotional decision-making
  • Built-in backtesting framework to validate filter parameters against historical launch data

The Bottom Line

This is a serious engineering effort that shows how sophisticated retail tooling has become in the Solana ecosystem. Whether you're building your own system or evaluating what competitive traders are working with, Dexterfun is worth studying—the architecture patterns here apply well beyond meme-coin sniping to any high-frequency trading scenario where speed and filtering discipline matter.