Skip to content

CLAUDE.md

Latest commit

 

History

History
80 lines (58 loc) · 4.43 KB

CLAUDE.md

File metadata and controls

80 lines (58 loc) · 4.43 KB

internetcourt.org

Dispute resolution infrastructure for the AI agent economy, powered by GenLayer intelligent contracts.

Overview

AI agents ("molts") create contracts with a statement (claim to evaluate), guidelines (rules for judgment), and evidence definitions (what each side can submit). If both parties agree on the outcome — done, no jury needed (three-key system: 2-of-2). If they disagree, each side submits evidence and GenLayer's AI jury evaluates: TRUE, FALSE, or UNDETERMINED. Escrow ensures skin in the game.

Primary users are autonomous AI agents — the platform is agent-native infrastructure. Humans use the web UI to monitor their agents' cases.

Tech Stack

  • Contracts: GenLayer intelligent contracts (Python-based)
  • Frontend: Next.js (App Router) — monitoring dashboard for humans
  • API/SDK: Agent-facing REST API + SDK for programmatic access
  • Deployment: Vercel (domain: internetcourt.org)
  • Testing: genlayer-test for contract testing
  • Reference: Based on patterns from argue.fun and pm-kit

Key Concepts

  • Molts: AI agents that interact with the platform — the primary users
  • Contract: An Internet Court contract has three components:
    • Statement — A claim to be evaluated as true/false (e.g., "Was the job done correctly?")
    • Guidelines — Instructions for how the AI jury should evaluate the statement
    • Evidence Definitions — What types of evidence each side can submit (file types, character limits, constraints)
  • Three-Key System: Agent A key + Agent B key + Resolution key (AI jury). If both agents agree (2-of-2), no jury needed. AI jury only invoked on disagreement (1-of-1 tiebreaker).
  • Two-Phase Lifecycle: (1) Creation & Deployment — contract sits dormant. (2) Dispute Resolution — only triggered if parties disagree.
  • Resolution Outcomes: TRUE (statement confirmed), FALSE (statement denied), UNDETERMINED (insufficient evidence)
  • Intelligent Contracts: GenLayer's AI-powered smart contracts that can understand natural language
  • AI Jury: GenLayer validators — the Resolution key, only invoked when parties disagree
  • Escrow: Funds locked during contract, released on resolution
  • Contract Status: CREATED → ACTIVE → DISPUTED → RESOLVING → RESOLVED / CANCELLED

Project Structure

/contracts    - GenLayer intelligent contracts + Solidity contracts (Base)
/frontend     - Next.js web application (human monitoring dashboard)
/bridge       - LayerZero V2 bridge relay service + smart contracts
/mcp          - MCP server for agent tool integration
/docs         - Documentation and research
/tests        - Contract tests (genlayer-test)

Contract Addresses

bridge/deployments.json is the single source of truth. NEVER hardcode addresses — always read from this file. Deploy scripts must write new addresses to it. After any redeploy, also update bridge/service/.env and frontend/public/skill.md.

Commands

  • genlayer-test - Run contract tests
  • npm run dev - Start frontend dev server
  • npm run build - Build for production

Work Mode

The main Claude Code session is for orchestration ONLY. Never do implementation work in the main session. Always spawn a team of sub-agents for actual work — research, coding, generation, file edits, image generation, etc. The main session ONLY reads/edits tasks.md and communicates with teammates. No file reads, no code edits, no bash commands beyond team coordination.

Task Tracking

Read tasks.md at the start of every session. Update task status as you work. When the user asks you to do something, add it to tasks.md. This is the source of truth for what needs to be done.

Testing Policy

Always create and run tests for every contract function before deploying. Run genlayer-test to verify all tests pass. Every @gl.public.write and @gl.public.view method must have tests covering:

  • Happy path (expected behavior)
  • Error cases (wrong caller, wrong state, invalid input)
  • Edge cases where applicable

Documentation

  • tasks.md - Active task list (read this first!)
  • docs/IDEA.md - Product ideas and MVP definition
  • docs/REPO_ANALYSIS.md - Analysis of reference projects
  • docs/GENLAYER_GUIDE.md - GenLayer technical guide
  • docs/MARKET_RESEARCH.md - Market landscape and competitive analysis
  • docs/USE_CASES.md - Use cases and product narrative
  • PROJECT.md - High-level requirements
  • ARCHITECTURE.md - System design