Skip to content

SUBMISSION_DESCRIPTION.md

Latest commit

 

History

History
106 lines (70 loc) · 3.89 KB

SUBMISSION_DESCRIPTION.md

File metadata and controls

106 lines (70 loc) · 3.89 KB

Submission Description

One-line summary

Arc Agent Research Marketplace lets one manager agent split a user request into specialist microtasks, pay each specialist in USDC on Arc, and resell the final report through a premium x402 unlock.

Track

Agent-to-Agent Payment Loop

Problem

Most AI products still treat agents like internal software features instead of independent economic actors. That breaks down when a workflow needs:

  • a specialist researcher
  • a separate verifier
  • a summarizer or evaluator
  • transparent machine-to-machine payment between them

Traditional chains make high-frequency micro-settlement too expensive and too noisy for this model.

Solution

We built a marketplace where:

  1. a manager agent turns one research brief into role-specific microtasks
  2. specialist agents handle research, factCheck, and summary
  3. every task is individually budgeted and settled in USDC
  4. agent identity, validation, and reputation are written on Arc via ERC-8004
  5. job execution is settled on Arc via ERC-8183
  6. the resulting report can be previewed for free and unlocked through a premium x402-compatible paywall

Why Arc

Arc fits this use case because the product only works if per-action settlement stays cheap enough for machine-to-machine commerce.

Arc gives us:

  • stablecoin-native payments in USDC
  • fast enough execution for repeated agent jobs
  • credible economics for sub-cent or near-sub-cent actions
  • a clean narrative for the emerging agentic economy

Circle + Arc stack

  • Circle Developer-Controlled Wallets for wallet creation and signing
  • Circle Developer Console for wallet and transaction operations
  • Arc Testnet for execution
  • ERC-8004 for agent identity, validator feedback, and validation
  • ERC-8183 for job creation, funding, submission, and completion
  • Nanopayments / x402 for pay-per-report monetization
  • Next.js for the demo dashboard and operator tooling

Live product proof

The shipped demo now includes:

  • a homepage with recent live Arc jobs and explorer links
  • a browser launcher for new jobs
  • judge mode with one-click autoplay demo
  • a premium teaser and paywall layer for report resale
  • economics UI that explains why the model breaks on traditional gas-heavy rails

The environment has already crossed the 50+ transaction threshold required by the hackathon. The homepage shows the latest visible portion of this onchain activity with explorer-ready tx links.

What judges should notice

  • This is not only a mock frontend. Judges can trigger a fresh live job from the browser and watch it land in recent runs.
  • The same workflow has both settlement logic and a monetization layer, which makes the business model much stronger.
  • The margin calculator explicitly shows why Arc is the right rail for this kind of per-action agent commerce.

Circle Product Feedback

Products used:

  • Arc
  • USDC
  • Circle Wallets
  • Circle Developer Console
  • Nanopayments / x402-compatible premium access

Why we chose them:

  • We needed programmable wallets, low-cost settlement, and a realistic micropayment path for agent-to-agent work.

What worked well:

  • Circle wallet setup made the programmable signing flow possible.
  • Arc made low-value specialist jobs and premium unlocks economically believable.

What could be improved:

  • A clearer hackathon-first setup flow for entity secret registration and wallet bootstrapping would reduce onboarding friction.
  • More official end-to-end examples combining Arc, Circle wallets, and x402 in one app would make shipping faster.

Recommendations:

  • Publish a first-party starter that bundles browser-triggered jobs, Circle wallets, Arc settlement, and x402 resale in one reference implementation.

Next expansion

  • batch multiple jobs from one client prompt
  • persist reports and metadata into durable storage
  • add live LLM outputs for each specialist role
  • add richer analytics for agent reputation and job throughput