Vincent

Vincent is an end-to-end system that lets autonomous AI agents access trustworthy, verifiable data — no centralized APIs required.
Here's how it works: A Chainlink CRE workflow runs on a schedule, pulling in market data and feeding it to an LLM to generate trading signals (BUY, SELL, or HOLD with a confidence score). The workflow then attests each signal on-chain, creating cryptographic proof that it came from a verified source and hasn't been tampered with. On the consumer side, an autonomous Python agent pays for access via x402 micropayments in USDC, proves it's human through World ID, and then uses those attested signals to make trading decisions with LangChain.
The core problem Vincent solves: today's AI agents either scrape unreliable data or blindly trust centralized APIs. Vincent flips that — signals are attested on-chain so agents can verify authenticity before acting. World ID keeps bots out, and x402 ensures you only pay for the data you actually use.

1. CRE Workflow — Written in the Chainlink workflow spec. Fetches external market data, calls OpenRouter for LLM inference, generates a signal, and attests it on Base Sepolia. Stores results in Supabase for the frontend.
2. x402 Server — Node.js/Express backend. Wraps the signal API with x402 payment middleware. Also handles World ID verification (v4 RP signature flow) and issues access tokens.
3. Frontend — React + Vite + Tailwind. Shows live signals, World ID gate, and agent activity. Connects to Supabase for real-time updates.
4. Agent — Python script using LangChain. Autonomously fetches signals, pays via x402, and logs decisions back to Supabase.