Hackathon agent for on‑chain quant
A hackathon project called Mercury used Nous Research’s Hermes agents to run real‑time on‑chain quant strategies, fraud detection and cashflow dashboards — the demo ties autonomous agents to trading telemetry and analytics shared. It’s an early example of agentic tooling applied to on‑chain trading workflows and operational monitoring.
Mercury — “Blockchain Cash Flow Analyzer” is published as a Hermes Agent skill for multi‑chain transaction fetching, period‑by‑period inflow/outflow breakdowns, fraud detection, AI trading‑strategy generation and a live interactive dashboard. (github.com) The repo lists explicit multi‑chain support for Ethereum, Base and Solana and requires API keys for Etherscan, Basescan and Helius to fetch on‑chain transactions. (github.com) Mercury implements six named fraud pattern detectors — dust attacks, rapid transfers, mixer interaction, address poisoning, dormant activation and high fan‑out — as separate analytic checks in its pipeline. (github.com) The project ships an AI trading strategy component that generates “trading insights based on wallet behavior patterns” and a three‑layer WebGL+Canvas+HTML dashboard with live particle animation, stat cards, gauges, charts and a risk grid. (github.com) Those features run as Hermes subagents in parallel via the skill’s use of Hermes’ delegate_task pattern, while Hermes Agent provides persistent memory, an auto‑creating skills system and the ability to run agents on low‑cost VPS or cloud hosts. (github.com) The Mercury README demonstrates an end‑to‑end agent workflow: a single Hermes command like “Analyze wallet 0x…” triggers automatic chain detection, parallel transaction fetching, simultaneous cash‑flow and fraud analysis, generation of trading signals and launch of the interactive dashboard. (github.com)
