Back
Explore EventsExplore ProjectsMy Projects
SCOUT - Verifiable Multi-Agent Intelligence — screenshot 1
SCOUT - Verifiable Multi-Agent Intelligence — screenshot 2
SCOUT - Verifiable Multi-Agent Intelligence — screenshot 3
SCOUT - Verifiable Multi-Agent Intelligence — screenshot 4
SCOUT - Verifiable Multi-Agent Intelligence — screenshot 5

SCOUT - Verifiable Multi-Agent Intelligence

Autonomous AI intelligence, cryptographically proven on-chain.

Agentic Loops Hackathon S1: Shanghai Edition

Links

Repository

github.com/Tasfia-17/scout.git

Website

scoutweb1.vercel.app

Demo video

Not set

Team

1 member
  • TA

    Tasfia Chowdhury

    owner
    Owner

Overview

Category: Web3/Blockchain

SCOUT is a verifiable intelligence platform that deploys a swarm of autonomous AI agents to perform complex research tasks. Given a single goal, SCOUT orchestrates multiple agents that browse the web in parallel, synthesizing their findings into a comprehensive, multimodal intelligence package in under 90 seconds. The output includes a written brief, a personalized outreach email, a spoken audio summary, and a visual storyboard.

The system's architecture is built on a Python and FastAPI backend, which manages agent lifecycles and communicates with the frontend via WebSockets. Each specialist agent, powered by an LLM like Qwen3-8B, uses Playwright for headless browser automation. What sets SCOUT apart is its deep integration of Web3 technologies for cryptographic verification. Each agent possesses a unique on-chain identity established using the ERC-8004 standard on the Taiko L2 network. Every action an agent performs is cryptographically signed with its identity, creating an immutable, tamper-proof audit trail.

Furthermore, SCOUT implements the x402 protocol, generating a real micropayment receipt for every API call, which adds an economic layer to the audit trail. Agents communicate and coordinate in real-time using an in-memory message bus based on the Google A2A protocol. This combination of high-speed, parallelized AI research with a robust, on-chain cryptographic proof system addresses the critical trust and verifiability gap in autonomous systems, making it suitable for enterprise and high-stakes applications.

Key features:

  • AI Agent Swarm: Deploys multiple autonomous AI agents in parallel to browse the web and collaboratively research any user-defined goal, completing tasks in under 90 seconds.
  • On-Chain Identity: Establishes a persistent, verifiable identity for each agent using the ERC-8004 standard on the Taiko L2 blockchain, enabling true agent accountability.
  • Cryptographic Audit Trail: Every action an agent takes is cryptographically signed using its on-chain identity, creating an immutable, tamper-proof record of its entire research process.
  • Verifiable Micropayments: Utilizes the x402 protocol to generate a real micropayment receipt for every API call, providing a granular economic ledger of agent activity.
  • Multimodal Output: Synthesizes research into a comprehensive intelligence package, including a written brief, a personalized outreach email, a spoken audio summary, and generated images.
  • Agent-to-Agent Comms: Implements the Google A2A protocol via an in-memory message bus, allowing agents to coordinate and share findings in real-time during a mission.

Tech stack: Python, FastAPI, Uvicorn, Playwright, Web3.py, Websockets, Aiohttp, Requests, Eth-account, HTML, Qwen3-8B, Google A2A Protocol, ERC-8004, Taiko L2, EIP-712, x402 Protocol, EIP-3009, MetaMask, Coinbase, Orpheus TTS, Flux Schnell, Aave V3, Base Sepolia

ExploreProjectsMine