SwarmFi - Ai Agent

SwarmFi is a fully autonomous DeFi trading system powered by a swarm of four specialized AI agents that operate without any central server or human intervention.

Each agent has a defined role: the Researcher analyzes market conditions using AI inference, the Strategist generates trading strategies, the Risk Guard verifies trust and safety using ENS-based identity validation, and the Executor performs real onchain trades via KeeperHub.

All agents communicate over Gensyn AXL using real peer-to-peer messaging, ensuring there is no centralized coordination layer. Shared state and memory are stored using 0G Storage, allowing agents to asynchronously collaborate through verifiable data blobs.

Trust is enforced at the protocol level through ENS, where each agent must prove its identity before participating in decision-making. This prevents rogue or malicious agents from influencing execution.

The result is a resilient, decentralized, and self-organizing financial system that can analyze, decide, verify, and execute trades completely autonomously — demonstrating a new paradigm for AI-native onchain coordination.

SwarmFi is built as a modular multi-agent system where each agent runs as an independent process and communicates via Gensyn AXL’s P2P messaging layer. Instead of direct function calls, all communication is done through real network messages routed via a local AXL node, making the architecture fully decentralized.

0G Storage is used as a shared memory layer. Agents upload and retrieve structured data (market analysis, strategies, execution logs) using root-hash addressing, enabling trustless data exchange across agents.

AI inference is handled via 0G Compute (qwen3 model), with fallback to external APIs for reliability. This powers market analysis and strategy generation.

ENS is deeply integrated for trust verification. Each agent has a registered ENS subname, and the Risk Guard checks onchain text records before approving any action. This creates a real trust-gating mechanism, not just identity labeling.

Execution is performed through KeeperHub’s API, where the Executor agent submits finalized actions onchain. A live dashboard built with Next.js and Upstash Redis visualizes real-time swarm activity.

A particularly notable aspect is that the system continues functioning even if individual agents fail, due to its decentralized P2P design.