AXL Agent Swarm

AXL Swarm — AXL Agent Swarm is a decentralized system of autonomous AI agents that collaborate peer-to-peer using Gensyn’s Agent eXchange Layer (AXL). The project consists of three specialized agents — News Researcher, Analyst, and Strategist — each running as a separate process on its own AXL node. After a single user query (e.g., “Top cybersecurity threats in Web3 right now”), the agents autonomously research information, analyze signals and risks, and synthesize actionable intelligence without any central server or message broker.All communication is encrypted and routed through AXL, demonstrating real peer-to-peer agent coordination. A Streamlit dashboard provides live visibility into the agent pipeline, current query, and final alpha recommendation.This project showcases the power of decentralized intelligence — agents that can discover, debate, and deliver high-quality insights in a fully distributed manner. It meets all hackathon requirements: separate AXL nodes, meaningful AXL usage, no central infrastructure, and clear utility in the Web3/AI space.Built from scratch during the Gensyn AXL Hackathon.

I built AXL Swarm — AXL Agent Swarm from scratch as a complete beginner during the Gensyn AXL Hackathon. The goal was to create a truly decentralized multi-agent system where three AI agents (News Researcher, Analyst, and Strategist) collaborate autonomously using only Gensyn’s AXL for peer-to-peer encrypted communication.Architecture:3 separate AXL nodes running as independent processes (each with its own config, private key, and API port: 9002, 9003, 9004) 3 Python agents, each connected to its own AXL node via simple HTTP calls (/send and /recv) Messages are passed as JSON with X-Destination-Peer-Id header for routing No central server, no message broker — everything goes through the AXL mesh

Tech stack:Gensyn AXL (Go binary built from source in WSL) — core P2P layer for discovery, encryption, and routing Python 3.12 + Groq (llama-3.3-70b-versatile) for agent intelligence Streamlit for the live dashboard showing query, pipeline flow, and final alpha File-based logging (swarm_log.txt) to sync real-time activity between agents and dashboard

Notable implementation details:Agents run as separate processes and communicate exclusively via localhost HTTP to their AXL node Simple polling loop in Analyst and Strategist (every 2 seconds) to receive messages Logging system I added so the dashboard can show live flow (News → Analyst → Strategist) Overcame multiple setup challenges: Windows → WSL migration, AXL build issues, API key limits, and Groq model deprecation

The hacky but effective part was using basic HTTP polling + file logging instead of complex message queues — it kept everything minimal and hackathon-friendly while fully leveraging AXL’s peer-to-peer capabilities.This project demonstrates real decentralized agent orchestration: one user input triggers autonomous collaboration across separate nodes with no central coordinator.