4663scan

A block explorer for Robinhood Chain (Arbitrum Orbit L2, chain ID 4663) — built to index the chain's live tip in real time and to treat tokenized stocks as first-class, not as an afterthought bolted onto a generic ERC-20 list.

Live at 4663scan.io · API docs

No live screenshot is included here — this README was written without access to a browser to capture one. Visit the link above to see it.

What makes it different

• Verified stock tokens, with counterfeit detection. Robinhood's tokenized-stock tokens are BeaconProxy contracts deployed by a single on-chain StockFactory. Anything on this chain that merely looks like a stock token by name ("Apple • Robinhood Token") is worthless as a signal — a copycat can set its own name to whatever it wants. So 4663scan never trusts a name match: every candidate is verified by asking the factory itself whether it deployed that exact address, and only a token the factory vouches for is ever labeled as real. Lookalikes that fail this check are indexed like any other ERC-20 — never shown as verified.
• A live-tip home feed, not a lagging index. Free RPC can't sustain indexing a ~100ms-block chain in real time forever, so most explorers either fall behind quietly or show a stale "last indexed" number as if it were current. 4663scan's home feed and list pages read the newest blocks straight from RPC on every request instead of waiting for the database to catch up — the tip is always the tip. Historical browsing still uses the indexed database, which is exactly what it's good at.
• Real contract pages. Bytecode identity, ERC-20/721 detection, proxy detection (EIP-1967/EIP-1822, resolved to the implementation), creator lookup, and verified source — checked against Sourcify and Blockscout — with tabs for source code, an interactive read-only console for any view function, and decoded event logs.
• A public, documented API. Every read endpoint the site itself uses is open to anyone: rate-limited, CORS-enabled, no signup, no key. See 4663scan.io/docs for the full reference with real example requests and responses.

Stack

• pnpm monorepo — apps/indexer, apps/api, apps/web, packages/shared
• Indexer — TypeScript + viem; WSS head subscription with HTTP polling fallback across a pool of independent RPC providers; head-first indexing with a resumable background backfill toward genesis; a single global rate limiter shared across every RPC consumer
• Database — Postgres 16 via Drizzle (embedded, user-space Postgres for local dev; DB_EMBEDDED=false points at a system server). Storage is tiered — full detail for a rolling hot window, compact "slim" rows kept forever, stock-token transfers never pruned — see docs/architecture.md
• API — Fastify, read-mostly REST + one SSE stream, aggressive response caching, global + per-route rate limiting
• Web — Next.js (App Router) + Tailwind, mobile-first

Quickstart

Requires Node ≥22 and pnpm.

pnpm install
cp .env.example .env      # defaults work out of the box for local dev
pnpm setup                # starts embedded Postgres, runs migrations, seeds labels
pnpm dev                  # indexer + api + web, concurrently

• Web: http://localhost:3000
• API: http://localhost:4663/v1/status (docs at http://localhost:3000/docs)
• Postgres: postgres://fourscan:fourscan@127.0.0.1:54663/fourscan

The Postgres data directory lives at ~/.local/share/4663scan/pgdata by default (override with PGDATA_DIR); it must be on a native Linux filesystem.

pnpm smoke runs an end-to-end smoke test against live mainnet.

Notes

• The indexer starts at the current chain head and works backwards in the background. Lookups older than the index start are lazily fetched from RPC and cached permanently once deep enough to be final.
• All RPC endpoints are configured via .env (see .env.example) — the indexer never exceeds INDEXER_MAX_RPS requests/second in total, split across every configured provider.

License

GNU AGPL v3. It's copyleft specifically because this is a network service: the AGPL's key difference from plain GPL is that running a modified version of this code to serve other users over a network counts as distribution, so anyone who stands up a competing instance from a fork has to publish their changes too, not just keep them behind their own API. A permissive license (MIT/Apache) wouldn't require that — anyone could take this, modify it, and run a closed competing explorer with no obligation to share anything back. AGPL doesn't stop someone from forking and running a competitor (nothing can, once code is public) — it just means they have to do it in the open.