Skip to content

Loop Dune

On-chain analytics for Loop Protocol's lending markets. Every pool and CDP vault is read straight from contract state on an hourly UTC grid since deployment, stored in Postgres, published to Dune, and shown on a dashboard.

A 75-second tour of the dashboard: overview, market tabs, the stress test, the data page and the API.

  • Dashboard

    Five analysis tabs per market (Overview, Yield, Flows, Risk, Rate model), an interactive stress test, a data-health page and a JSON/CSV API.

    Feature tour

  • Data

    One long fact table of exact uint256 values plus SQL views for every metric, mirrored as public Dune tables with ready-made queries.

    Data model · Dune

  • Indexer

    Python, Multicall3, a registry verified on-chain, idempotent and self-healing runs. One call per chain per hour.

    Architecture

  • Hosting

    Managed (Supabase + GitHub Actions + Vercel) or a single VPS with HTTPS, continuous deployment and encrypted, restore-tested backups.

    Choose a setup

At a glance
Markets lpETH and lpUSD on Ethereum; lpBNB and lpBTC on BNB Chain
Coverage 4 pools, 4 staking vaults, 16 CDP vaults, 146 series, from each pool's deployment block
Resolution hourly (UTC), with daily rollups
Outputs Postgres views · public Dune tables · Next.js dashboard · JSON/CSV API
Status The Loop frontend has been sunset; the contracts are still on-chain, so the data can still be read, audited and charted.

How it works

flowchart LR
  rpc[(Archive RPC<br/>Ethereum · BNB Chain)] -->|Multicall3, 1 call/hour/chain| idx[loop-indexer]
  idx -->|idempotent upserts| pg[(Postgres)]
  pg -->|append-only| dune[(Dune tables)]
  pg --> dash[Dashboard + API]
  • Registry verified on-chain. Tokens, decimals, deployment blocks and each pool's vault list are re-checked against the chain (loop-indexer registry verify).
  • One UTC grid for every chain. "State as of 14:00 UTC" means the last block at or before 14:00 on each chain, so Ethereum and BNB Chain series join row for row.
  • Exact and honest. Raw uint256 is kept alongside scaled values. A reverting call is recorded as missing, never as zero.
  • Self-healing. Every run fills every missing hour since deployment. Re-runs are no-ops, and outages repair themselves on the next run.

What the data shows

The indexer surfaced things the protocol's own accounting hid:

  • A vault borrowed 120k USDC against dust. The syrupUSDC vault's oracle reports prices scaled 1e30 instead of 1e18. Debt peaked at 124k USDC; 4,000 USDC is still outstanding.
  • Lenders took losses. The lpETH share price peaked at 1.0155 (5 Feb 2025) and fell to 1.0012 by mid-2025, a ~1.4% loss of deposit value.
  • The rate model was retuned. At 75–85% utilization, lpETH charged ~21% APR before June 2025 and ~7% after: two separate curves on the Rate model tab.
  • An oracle collapsed. sdeUSD's price fell 90% in one day (7 Nov 2025) and its oracle has reverted ever since.

More in the data model.

Quick start

Requirements: uv, Node 20+, and Docker (or any Postgres ≥ 14).

git clone https://github.com/guilyx/loopscan && cd loopscan
cp .env.example .env          # add archive RPC URLs
make setup                    # Python + Node dependencies
make db migrate               # local Postgres + schema
make backfill                 # daily snapshots since deployment
make run                      # fill in hourly resolution
make export-static dashboard  # http://localhost:3000

No database or RPC handy? The dashboard runs on the committed snapshot: cd dashboard && npm ci && npm run dev.