I build software that AI agents can trust: APIs, automations and payment rails that only charge when the work is actually done.
If the API doesn't keep its promise, the agent doesn't pay.
Pay-on-validation for x402: the AI agent's payment is verified but only settled after the response is judged against the seller's published contract (schema → objective rules → AI judge). Built for the MoonPay PayBox hackathon.
- ✅ Settled with real USDC on Solana mainnet and Base
- 🛡️ 40 paid calls, 19 broken promises blocked, $0 charged for failures
- 🧾 Promises written in plain language, compiled into deterministic / heuristic / probabilistic checks
- 🔏 Every verdict signed with Ed25519 and verifiable offline
- 🔌 MCP plugin for PayBox agents
Repo · Live demo: try to break it · Video
TypeScript Node.js Hono x402 viem Solana Kit Claude API MCP Zod
| Languages | Python · JavaScript · TypeScript · SQL |
| Web | React · React Native · Node.js · Hono · REST APIs |
| Data | MySQL · PostgreSQL · Pandas |
| AI & Web3 | Claude API · Claude Code · MCP · x402 · Base · Solana |
| Tools | Git · GitHub · Postman · Docker · VS Code |
Also learning: cybersecurity (application security, networks, ethical hacking).
- Taking Tribunal from hackathon to production: one-line integration for any x402 API, on-chain escrow and a public trust score for agent-payable services
- Open to remote roles and collaborations on AI agents, payments and automation
💬 Ideas, feedback or want to break Tribunal? Open an issue or DM me on X.