Cred Protocol
Book a demo

Real borrowing history. One score you can defend.

Cred reads every repayment, liquidation and new borrow on Aave, Compound and Morpho, on Ethereum and Base, adds how the wallet behaves across ten chains, and turns it into a 300 to 1000 score with factor-level reasons your credit team can sign off on.

Building an AI agent? Get an API key and score wallets over MCP or x402.

  • ETH, borrowing history and wallet behaviour
  • Base, borrowing history and wallet behaviour
  • ARB, wallet behaviour
  • OP, wallet behaviour
  • POL, wallet behaviour
  • AVAX, wallet behaviour
  • BNB, wallet behaviour
  • Celo, wallet behaviour
  • Linea, wallet behaviour
  • HOOD, wallet behaviour
0x7a3f…c91e · sampleandromeda_1.4
862Very Good
Borrowing
Excellent
Health
Very Good
New Credit
Good
  • Repay · Ethereum · 950 USDC
  • Borrow · Ethereum · 1.2 WETH
Lenders and wallets
  • 3Jane
  • Teller
  • Atlendis
  • MetaMask
  • Celo
Backed by
  • Robot Ventures
  • DCG
  • Alliance
  • Volt Capital
  • Orange DAO

Score it, read it, watch it.

Three APIs over the same lending record: a score for the decision, a report for the file, and webhooks for the moment something changes.

GET /api/v2/score/address/0x7a3f…c91e
862Very Good
andromeda_1.4 · decile 7 · 6 factors

Credit Scoring

The Cred Score ranks wallets by their risk of liquidation, default or delinquency. Wallets without borrowing history are still scored from their onchain behaviour.

Score any wallet
Net worth
$48,210
Outstanding debt
$12,400
Debt to collateral
0.31
Active chains
7 of 10

Credit Reporting

Multi-chain reports on wallet composition by asset, chain and protocol, with the lending metrics an underwriter reads: debt, collateral and their ratio.

Read a credit report
  • liquidation· Aave V2 · Ethereum
  • transfer· Base · 25,000 USDC
  • borrow· Morpho Blue · Base
  • POST → your webhook · 200

Credit Monitoring

Signed webhooks for the events that change a credit decision: liquidations, new borrowing, repayments and transfers on the wallets you watch.

Set up credit alerts

Why the score holds up in committee.

Every response names the model that produced it, every model change is published, and missing data is reported as missing, never as good or bad.

  1. andromeda_1.4liveNov 2026Morpho Blue repayments and liquidations
  2. andromeda_1.3Sep 2026New Credit: recent originations and their pace
  3. andromeda_1.2Sep 2026Compound V2 and V3 repayments and liquidations
  4. andromeda_1.1Sep 2026Aave V2 and V3 repayments and liquidations
  5. andromeda_1.0Dec 2024Wallet behaviour across chains

Every score names its model

Responses carry model_version. A new version adds inputs; the weights before it stay fixed, so a score moves for a reason you can point to.

Read the model release notes
compound_v3 · Base · feeds_score: true
new_protocol · Ethereum · feeds_score: false
Ingested and reported, not yet scored

New inputs run in shadow first

A newly indexed protocol is read and reported on every wallet before it is allowed to move a score.

coverage: "partial"
unavailable_chains: [59144]
No chain answered? 503, not a score

Gaps are reported as gaps

When a chain's data provider cannot answer, the report says which chain, instead of scoring the wallet as if it had no history.

How the score is checked

No published default rates: we would rather show you the method and test it on your own book.

Test it on wallets you know
Release-gated behaviour tests
Before a model version ships it must rank a fixed set of borrower archetypes the way an analyst would: a defaulter below a clean profile, a recovered defaulter above a serial one, an empty wallet lowest of all.
No borrowing record, no top band
A wallet that has never borrowed is scored from its behaviour but capped at 839, so it never outranks a proven borrower.
Your book is the real test
On the call we score wallets whose outcomes you already know, so you can judge the ranking against your own history.

The same score, one API call away.

Score, report and monitor wallets from your own code or your agent. Build against the sandbox, then drop /sandbox from the path for production.

  • A sandbox with mock wallets for every endpoint, at no Cred Unit cost
  • TypeScript and Python SDKs, with Express, Hono, Next.js and FastAPI adapters
  • An MCP server, so AI agents can score wallets directly
  • Pay per call over x402 or MPP, no account needed
Get an API key
GET /api/v2/score/address/0x7a3f…c91e?include_factors=true
{
  "address": "0x7a3f…c91e",
  "model_version": "andromeda_1.4",
  "score": 862,
  "decile": 7,
  "range": "very_good",
  "factors": [
    { "label": "Borrowing History", "rating": "excellent", … },
    { "label": "Wallet Health", "rating": "very_good", … },
    { "label": "New Credit", "rating": "good", … }
  ]
}

From the blog

All articles

See your own wallets scored.

A 30-minute call with live scoring on wallets you choose, and the factor reasons behind every number. No commitment.

Building an AI agent? Get an API key