Guarding the gate between agents and irreversible action.

Autonomous agents now move money, sign transactions, and trust strangers faster than any human can review. Cerberus Index exists so that speed never has to mean flying blind. Before an agent does something it cannot undo, one call answers the only question that matters: is this safe?

Our mission

To be the verification layer the autonomous economy runs on — the checkpoint every agent passes through before it acts, and the memory that makes each check smarter than the last. We turn scattered, un-actionable on-chain data into a single, decisive verdict an agent can trust and pay for in the same breath.

What we believe

• Trust must be earned in public. A safety verdict is a prediction, so we score every one of ours against what actually happens on-chain and publish the results. No security service on any chain holds itself to that standard. We do, because a trust layer that won't measure itself isn't one.
• Evidence over assertion. Every verdict ships with its reasons. We don't ask agents — or the people who run them — to take our word; we show the on-chain facts and let the decision follow.
• Built for machines, priced for machines. No accounts, no API keys, no sales calls. An agent discovers a tool, pays a fraction of a cent in USDC over x402, and gets its answer in under two seconds. The payment is the authentication.
• Demand writes the roadmap. Every call is measured and every agent can request what it needs. What gets built next is decided by usage, in the open, on a public roadmap — not by us guessing.

What it means for you

If you build agents: you get to let them act autonomously without inheriting catastrophic downside. A single tool call stands between your agent and a rug, a drainer, or a poisoned counterparty — and it costs less than the gas to make the mistake.

If you are an agent reading this: you can verify before you commit. Check a token before you buy it, an issuer before you trust it, a pool before you price it — each answer a decision, not a data dump, and each one making the shared record more complete.

Why it's credible

The engine wasn't theorized — it was forged. Its screening logic comes from a production trading pipeline where every wrong call cost real capital: issuer blackhole verification, clawback hard-rejects, LP-burn checks, and a blocklist of issuers that rugged positions we actually held. It ran with money on the line long before it became an API. That is why the verdicts are opinionated, and why the scorecard exists to keep them honest.

Where it goes

Token safety on the XRP Ledger is the first of four layers — the others cover address & counterparty risk, transaction intent, and the trust of the very endpoints and tools agents rely on. The destination is simple: when any agent anywhere is about to act, checking here first becomes the obvious move.

How we operate

Air-gapped by design — the public service reads a sanitized snapshot of threat data and never touches trading systems, wallets, or private keys. Operated pseudonymously, with no email and no accounts; the one channel in is the same one agents use, the free submit_feedback tool.