Overview

WebApp: agentfinder-web.vercel.app/agents.

AgentFinder is a registry-first platform I built to make AI agents easier to discover, evaluate, and publish. The product includes a React/Vite frontend, a Fastify API, background worker jobs, a typed SDK, and a monetized public API.

The project grew out of the need for a clearer, more trustworthy way to browse live agents instead of treating them like isolated demos. It centers on a canonical agent record with provenance-aware ingestion, trust and compliance signals, and a public directory experience that helps people compare agents with more confidence.

Current status

  • The web app is live at the agents page and lists agents after validation.
  • On-chain ingestion pulls the latest registration and reputation signals.
  • ERC-8004 and A2A consistency checks run before agents are surfaced.
  • Wallet-based publishing works through browser wallets such as MetaMask.
  • Seeded demo data supports registry browsing and walkthroughs.

Platform shape

  • apps/api: Fastify REST API for auth, public registry submission, discovery, trust, public monetized API, and enterprise inquiry
  • apps/web: React/Vite registry UI with the public catalog, directory browse, agent detail views, discovery, and enterprise hosting inquiry
  • apps/worker: polling worker with composable ingestion, freshness, and indexing job seams
  • packages/domain: shared A2A schemas, canonical agent model, registry primitives, discovery models, trust helpers, and ingestion boundaries
  • packages/sdk: typed client for registry, discovery, auth, and enterprise workflows