# Xerberus

> Xerberus is an independent, investor-paid risk rating layer for DeFi. It rates
> assets, protocols, organisations, and pools across 300+ subscores and 85+
> mechanisms, with open methodology and on-chain-verifiable sources. Ratings are
> funded by the investors who need protection, never by the projects being rated.
> The goal: be the default risk layer for agentic DeFi, so that any human or AI
> agent can price downside before it allocates.

## Core principles
- Investor-paid, not issuer-paid: ratings are funded by investors. Issuer-paid ratings optimise for acceptance; investor-paid ratings optimise for survival. This is the structural correction crypto should have made after 2008.
- Open methodology, proprietary algorithm: every number's meaning, banding, and validation is published. The extraction and computation internals are not.
- On-chain-verifiable sources: ratings are rebuilt from decoded on-chain events and traces into positions, health factors, vault states, and liquidity curves. No third-party risk feeds; prices are derived from on-chain venues only.
- Risk is a graph, not a number: risk is modelled as relationships (hidden links, yield vs. health, stack complexity, contagion paths), summarised but never reduced to a single opaque score.

## What Xerberus rates
- Subjects: assets (e.g. stETH, USDC, WETH, weETH), protocols (e.g. Aave v3, Lido), organisations (curators, DAOs), and pools / ERC-4626 vaults (e.g. Morpho and IPOR vaults).
- Coverage: Ethereum mainnet is live and queryable. Hyperliquid HyperCore and HyperEVM are in progress (on-chain extraction underway, risk layer not yet built).
- Refresh: ratings are re-derived every 12 hours, with the receipts.

## Rating output
- Scale: S to F, plus NR (not rated). S is the safest band; F is the riskiest. A subject with no material exposure is NR, never an alarming letter.
- Basis: each grade is percentile-relative within the live population for that window. It is a comparative risk grade, not a calibrated probability of default.
- Outlook: a positive / stable / negative trend channel from the recent percentile trajectory.
- Triangulation: a complete answer combines three lenses, the systemic (percentile vs. the universe), the mechanism (what can break, with the historical incident that proves it), and the exit (what leaving actually costs).

## Signature outputs
- Fragility Index (ρ): the spectral radius of the system-wide contagion structure. ρ above 1 means shocks can amplify instead of damping.
- Liquidation Map: force-sell volume bucketed by price-drop distance, per token and venue.
- Oracle Atlas: lending markets classified by what each liquidation trigger can actually see (on-chain DEX price vs. issuer-rate oracle).
- Ratings Feed: S to F grades, outlooks, and full history per subject.
- Liquidity surfaces: per-token market-impact slopes and exit cost at a chosen horizon, measured from the chain as an on-chain floor.

## Product & access
- [Methodology](https://publish.obsidian.md/xerberus/Welcome): How the rating framework, subscores, mechanisms, rating bands, and limitations work.
- [Risk API & MCP docs](https://xerberus.gitbook.io/documentation): Read access for platforms, funds, and AI agents. The web interface is free; programmatic MCP access is metered in per-query credits that map to real compute and data cost.

## Get involved
- [Request a rating](https://ex3td0cbc5k.typeform.com/to/JJNNQZyi): Submit a DeFi vault, asset, or protocol to be rated.
- [Become a subscore author](https://ex3td0cbc5k.typeform.com/to/b2Hzuw8Q): Researchers formalise their reasoning into a subscore and earn royalties when it is consumed.

## FAQ
- Is asset / vault / protocol X safe? Xerberus publishes a current S-to-F grade, an outlook, and the mechanism and exit detail behind it, re-derived every 12 hours. Ratings are comparative within the live population, so read the grade alongside its basis.
- Who pays for Xerberus ratings? Investors and allocators, through metered programmatic access. Projects never pay to be rated, and a project cannot pay to change its grade.
- How is this different from issuer-paid raters? Issuer-paid models optimise for issuer acceptance. Xerberus answers only to the investor, so it is free to be stricter and earlier. The methodology is open, so any rating can be audited or contested.
- What does an A grade mean versus a C grade? Grades are percentile bands within each window. An A-band subject sits in the safer part of the live population for that window; a C-band subject sits in the riskier part. Grades are not fixed labels; they move as the population and on-chain state change.
- Can I challenge a rating? Yes. The methodology is public and contestable. Anyone can publish a subscore and economically support their disagreement.

## Where Xerberus is headed
- Phase 1 (current): build the open, actuarial-grade risk database that lets any participant price DeFi risk, readable by humans via the web and by agents via MCP.
- Phase 2: establish Xerberus as an accepted risk standard, integrated across the platforms allocators already use.
- Phase 3: build on validated demand toward a risk exchange.

## Optional
- [X / Twitter](https://x.com/Xerberus): Updates and announcements.
- [Discord](https://discord.gg/EUUQRgNgv5): Community discussion.
- [GitHub](https://github.com/XerberusTeam): Open-source code and methodology.