Pump.fun Auto-Trading Bot v2

A Solana automated trading bot for pump.fun tokens. Scans, scores, buys, and sells tokens automatically with built-in rug detection.

Architecture

┌──────────────────────────────────────────────────────────────┐
│                        BOT CORE (bot.py)                      │
│                                                              │
│  Scanner ──→ Enrich ──→ Rug Check ──→ Score ──→ Buy         │
│     ↑                                              ↓         │
│     │         ┌─────────────────────────────┐     │         │
│     └─────────│   Position Manager          │←────┘         │
│               │   (TP / SL / Trailing)      │               │
│               └────────────┬────────────────┘               │
│                            ↓                                 │
│                         SELL (Jupiter / PumpPortal)           │
└──────────────────────────────────────────────────────────────┘
         │                                           │
         ▼                                           ▼
    Dashboard (port 4001)                   Telegram Alerts

Key Improvements Over v1

Issue	v1 (Broken)	v2 (Fixed)
Name Resolution	Always "?"	3-source resolver (DexScreener → Pump.fun API → Helius DAS)
Sell Path	100% failure	Jupiter V2 API + PumpPortal fallback
Buy Failures (90%)	No pre-checks	Balance check + bonding curve check + graduation check
Duplicate Buys	No dedup	CA-based dedup (never buy same token twice)
Rug Detection	None	Multi-factor analysis (creator %, authorities, socials, name patterns)
Scoring	min=50, broken	min=72, working name/momentum/holder scoring
MC Range	$5K-$50K	$15K-$80K (avoids dead launches)
Position Staleness	None	Auto-close after 4 hours
Wallet Balance	Not checked	Pre-flight balance check

Files

File	Purpose
bot.py	Main orchestrator
config.py	All settings/thresholds
scanner.py	Token discovery (DexScreener + pump.fun)
token_resolver.py	Name/symbol resolution (fixes "?" issue)
scoring_engine.py	Multi-factor scoring (0-100)
rug_detector.py	Rug pull detection
trader.py	Buy/sell execution (PumpPortal + Jupiter)
position_manager.py	Position tracking & sell logic
data_store.py	Persistent storage
notifier.py	Telegram alerts
dashboard.py	Web UI (port 4001)

Quick Start

# 1. Copy and fill in your environment variables
cp .env.example .env
nano .env

# 2. Run the bot
chmod +x start.sh
./start.sh

Manual Start (without script)

pip install -r requirements.txt
python3 bot.py

Configuration

Edit config.py to tune:

• BUY_AMOUNT_SOL (0.03): How much SOL per trade
• MIN_SCORE (72): Minimum score to buy (higher = pickier)
• MIN_MC_USD (15,000): Skip tokens below this market cap
• MAX_MC_USD (80,000): Skip tokens above this (too late)
• STOP_LOSS_PCT (-35%): Hard stop loss
• TAKE_PROFIT tiers: Sell portions at +50%, +150%, +400%
• TRAILING_STOP: Activates at +25%, trails 12% below peak

How It Finds Good Coins

1. Scanner checks DexScreener + pump.fun every 8 seconds for new tokens
2. Resolver fetches real name/symbol (never buys unknowns)
3. Basic Filters eliminate tokens outside MC range, graduated, or too old/new
4. Rug Detector rejects tokens with:
   • Creator holding >15%
   • Top 10 wallets holding >60%
   • Active mint/freeze authorities
   • No socials (twitter/website)
   • Scam name patterns
5. Scoring Engine rates remaining tokens (0-100) on:
   • Momentum (30%): How fast MC is growing
   • Holders (25%): Distribution quality
   • Social (15%): Twitter/Telegram/Website presence
   • Liquidity (15%): Exit liquidity depth
   • Name (10%): Meme virality potential
   • Volume (5%): Trading activity

Only tokens scoring 72+ with rug score <50 get bought.

Dashboard

Access at http://YOUR_VPS_IP:4001

Shows: scanner stats, P&L, active positions, recent trades, wallet balance.

Telegram Setup

1. Message @BotFather on Telegram, create a bot, get the token
2. Message @userinfobot to get your chat ID
3. Add both to .env

VPS Deployment

# Run in background with systemd
sudo cp pumpbot.service /etc/systemd/system/
sudo systemctl daemon-reload
sudo systemctl enable pumpbot
sudo systemctl start pumpbot

# Check logs
journalctl -u pumpbot -f