Risk you can watch, not just read about
Stenion is an open-source, live risk-intelligence platform for Stellar and Soroban DeFi lending protocols. It reads each protocol directly from the chain and continuously scores how safe it is right now — then publishes exactly how it got that number.
Why continuous scoring
Two things dominate how people currently judge DeFi risk, and both miss the moment that matters:
- Static audits confirm the code was sound on a particular date. They say nothing about the pool becoming 95% concentrated in one asset, or an oracle drifting stale, weeks later.
- TVL trackers (DeFiLlama already does this well for Stellar) measure how much is at stake — not how risky the position is. Size is not safety.
Stenion's differentiator is the gap between those two: continuous, on-chain-derived risk scoring — collateral concentration, oracle staleness, admin-key control, liquidity depth, and utilization headroom — recomputed on an interval and timestamped so you always know how fresh a number is.
How the score is built
Every protocol is graded on the factors its category is scored on — five for a lending market, two for an AMM — each on a 0–100 scale where higher is safer. The overall safety score is a fixed weighted mean of those factors, so two scores mean the same thing only when the categories match, which is why the registry ranks each category separately. Wherever possible a factor is anchored to the protocol's own on-chain parameters — Stenion grades a pool against the line it set for itself, not against an arbitrary constant.
Adapters read data directly from Soroban RPC and Horizon — official, trustless Stellar infrastructure — so nothing is self-reported by the protocols being scored. The exact formulas are public in the methodology, and they're meant to be challenged.
Rules that cannot be bought
These are non-negotiable and enforced in the methodology itself.
Payment never affects the score
Protocols can pay for visibility, speed, or private tooling — never for a better number. Paid "Spotlight" placement is a visually separate, clearly labeled section. The real ranking is always free, public, and purely score-ordered.
AI never invents an assessment
Any AI feature only explains or summarizes the real underlying data. It never generates an independent risk judgment and never sets a score.
No fabricated numbers
When real data genuinely isn’t available for a factor, Stenion uses a clearly-flagged neutral baseline and says so — never an invented, plausible-looking value.
Open source, one adapter per protocol
Coverage grows DeFiLlama-style: each protocol is a small, open-source, PR-reviewed adapter reading directly from on-chain infra. Anyone can inspect, verify, or contribute one.
Honest about the limits
Stenion measures what it can read trustlessly on-chain, and is explicit about what it can't. Oracle scoring grades two things: how stale the worst price is, measured against that oracle's own publish resolution and maximum acceptable age rather than a Stenion constant, and whether the pool's price path actually bounds how far a single update can move it — including whether that bound is armed, since a configured limit with no stored baseline is inert. The factor takes the worse of the two, because a bounded stale price and a fresh unbounded price are both untrustworthy.
What it still doesn't measure is market depth — whether the bound is generous relative to how cheap the underlying market is to move. A thin market can be pushed within an allowed deviation. That is a real, recorded limitation, not a hidden one, and it is open on the roadmap rather than quietly assumed away. Every threshold is labeled as either anchored to an external/on-chain value or an unvalidated judgment call open to challenge.
It's built in the open by a solo developer in the Stellar ecosystem. If you think a threshold is wrong — especially if you're a protocol being scored — the methodology explains how to dispute it.
See it working
Live scores for the protocols Stenion tracks today.