Building a six-agent memecoin trading algorithm with Astra
The author says GPT-6 Astra and Fable 5.1 were used to build QuantumOrb, a six-agent Solana trading bot that made 521 trades in a month and +17.1 SOL, with code on GitHub.
I used GPT-6 Astra and Fable 5.1 to build a self-learning memecoin trading algorithm run by 6 AI agents I named it QuantumOrb 521 trades in a month, about 17 a day. Average trade size: 0.48 SOL. Result: +17.1 SOL 6 AI agents analyze new memecoins on Solana and learn to make trading decisions as one system More details about this bot on GitHub: https://github.com/VALIXxDaniro/QuantumOrb Here’s how it works: > SCOUT tracks new launches, bonding curve trades and swaps on PumpSwap. Records who bought, how much and at what price > CARTOGRAPHER looks for shared wallet funding sources, buys in the same bundle and links to the token’s creator. Linked buyers count as one group > JUDGE evaluates traders based on closed trades. Checks the consistency of their results and assesses their edge after accounting for fees, entry delays and sniper bots > QUANT calculates how a trade would fill on the curve or in a pool. Deducts fees, the priority fee, the buy’s price impact and the Jito tip. Determines the position size > GUARD checks holder concentration, linked groups and launch risks. Can veto a trade no matter how promising the chances of a price rise look > EXECUTOR models entries and exits: take-profit, stop-loss or a time-based exit. Currently operates in research mode At the center of the system is Orb, which holds three probabilities for each launch it evaluates: +60% is reached first −30% is reached first neither threshold is reached within 60 minutes Organic buying flow reinforces one scenario. Linked wallets, selling and concentration change the estimate. Every new signal updates the state Then Quant checks whether an entry makes sense after all costs, while Guard checks whether the risk is acceptable Every launch that passes Guard becomes a training example The system is still learning and currently trades only with virtual capital. Once it is trained and stable, I plan to give it real capital and make its trading visible through a public wallet