Professional AI-assisted trading framework for Pump.fun and Solana memecoin workflows.
This project upgrades a legacy Pump.fun automation concept into a modular Pump.fun trading bot with token scanning, AI validation, sniping, risk controls, and portfolio management.
This repository is designed for builders and advanced users who want to operate a Pump.fun bot or Pumpfun trading bot from a command line dashboard with configurable logic.
Core idea:
- Detect new launches quickly.
- Filter aggressively using AI + on-chain heuristics.
- Execute fast entries with priority fees.
- Manage risk with strict stop-loss/take-profit rules.
- Track results and continuously tune strategy.
- Monitors newly launched tokens from Pump API and on-chain program logs.
- Captures metadata including mint, creator, launch timestamp, and additional raw context.
- Persists scan history in local storage for audit and backtesting.
- Uses OpenAI to score each candidate token using structured prompt context.
- Combines AI output with hard filters:
- social footprint checks (optional)
- suspicious keyword detection
- mint/freeze authority heuristics
- early momentum proxy
- Standard output format:
score(0-100)reasoningaction(BUYorSKIP)
- Builds fast transactions with:
- compute unit limit instructions
- priority fee bidding
- simulation before sending
- Supports manual and auto-snipe flows.
- Designed for extension into full Pump.fun instruction encoding and Jito routing.
- Bundle module can batch multiple wallet actions in one execution cycle.
- Useful for advanced entries/exits and coordinated strategy execution.
- Toggle with
BUNDLE_ENABLED.
- Tracks open and closed positions with realized PnL.
- Supports stop-loss, take-profit, trailing stop configuration.
- Provides live CLI summaries for active positions and performance.
- Rotates proxies from
proxy_list.jsonfor outbound requests. - Helps reduce request concentration and API rate-limit pressure.
- Start Scanner
- Stop Scanner
- View Recent Scans / Tokens
- View Portfolio
- Manual Snipe (by mint)
- Exit Position
- Settings
- Exit
src/index.ts- bootstrapsrc/runtime.ts- wiring and dependency containersrc/config.ts- environment-driven configurationsrc/scanner.ts- launch discoverysrc/analyzer.ts- AI + heuristic scoringsrc/trader.ts- execution and transaction build flowsrc/bundler.ts- bundle abstraction layersrc/portfolio.ts- position lifecycle + PnLsrc/storage.ts- persistent local statesrc/proxy.ts- proxy rotationsrc/cli.ts- dashboard and operator commands
- Scanner event arrives with mint + launch metadata.
- Analyzer enriches data (on-chain checks + optional social checks).
- OpenAI scoring returns structured JSON decision.
- If score >= threshold and action is BUY, sniper engine executes with priority settings.
- Position opens and gets tracked in portfolio store.
- Risk rules (SL/TP/trailing/manual exit) manage lifecycle.
- Results are logged for future parameter tuning.
No trading bot can guarantee profit. Real performance comes from risk control + disciplined iteration.
Recommended framework:
- Use small position size first (
SNIPE_AMOUNT_SOL=0.01or less). - Keep strict stop-loss (
STOP_LOSS_PERCENTlow, e.g. 6-10%). - Start with high AI threshold (
AI_VALIDATION_THRESHOLD=80+) to reduce noise. - Enable social checks, but treat social signals as weak evidence unless confirmed on-chain.
- Track metrics weekly:
- win rate
- average R multiple
- max drawdown
- slippage/fill quality
- Increase size only after statistically meaningful positive expectancy.
- Use separate wallets per strategy archetype for cleaner analytics.
Practical strategy improvements:
- Add blacklist/whitelist creator wallet logic.
- Add time-based invalidation (skip if entry is late after launch).
- Add volatility-adaptive TP/SL.
- Add fail-safe cooldown after consecutive losses.
- RPC providers: Helius, Triton, QuickNode
- Bundle relays: Jito APIs / custom relay pipelines
- Data sources: DexScreener, Birdeye, social sentiment providers
- Notifications: Telegram/Discord/Slack alerts
- Persistence: SQLite/Postgres for advanced analytics
- Observability: structured log shipping + metrics dashboards
- Install dependencies:
npm install- Configure environment:
copy .env.example .env-
Edit
.envwith your keys and risk settings. -
Start CLI:
npm run devRPC_HTTP_ENDPOINT- Solana HTTP RPC endpointRPC_WS_ENDPOINT- Solana WebSocket RPC endpointOPENAI_KEY- OpenAI API keyBOT_KEY- main trading wallet private key (base58)EXTRA_WALLET_KEYS- optional comma-separated extra walletsJITO_AUTH_KEY- optional Jito auth keyPUMP_PROGRAM_ID- Pump.fun program idAI_VALIDATION_THRESHOLD- minimum score for auto-buySNIPE_AMOUNT_SOL- SOL size per entrySLIPPAGE_BPS- max slippagePRIORITY_FEE_LAMPORTS- fee priority for faster inclusionSTOP_LOSS_PERCENT- stop-loss percentageTAKE_PROFIT_PERCENT- take-profit percentageTRAILING_STOP_PERCENT- trailing stop percentageSCAN_INTERVAL_MS- scanner poll intervalBUNDLE_ENABLED- bundle mode toggleSOCIAL_CHECK_ENABLED- social checks toggleMAX_RECENT_TOKENS- local scan history capLOG_LEVEL- logging verbosity
- Never commit real
.envvalues. - Use a dedicated hot wallet with limited balance.
- Keep backups of strategy settings and trade logs.
- Validate all execution paths on devnet/small size before scaling.
- Monitor RPC errors and fail rates continuously.
This is a high-risk trading automation tool.
There is no guarantee of profits. You are fully responsible for losses, operational risk, and legal compliance in your jurisdiction.
pumpfu trading bot, Pump.fun bot, Pumpfun trading bot, Pumpfun comment bot, Pump.fun automation, Solana trading bot, Solana meme coin bot, token engagement bot, AI crypto bot, OpenAI trading bot, Pumpfun volume support, Pumpfun marketing bot, crypto community growth bot, on-chain token promotion bot, automated comment posting, Pump.fun CLI bot