Most trackers show price. Pharos shows risk.
Live reserve feeds, forward-looking depeg warnings, and public blacklist monitoring across a 268-asset core universe of stablecoins and cash equivalents: everything that could make a stable asset fail.
Active stablecoins
392
Every stablecoin makes two promises at once: that it will redeem at par, and that it can be redeemed at all. Pharos watches both across 268 core assets and 97 tracked variants, plus 27 stable-value investments, and the chains and reserves beneath them so that desks, treasuries, and researchers can read the peg the way navigators read weather.
Edited by TokenBrice · Engineered with Claude & Codex · MIT
Watching the peg.
Why it exists
The data needed to evaluate stablecoins is scattered, inconsistent, or buried behind paywalls. Worse, marketing claims rarely match reality: a "decentralized" stablecoin may rely entirely on centralized collateral, an issuer can freeze balances without warning, and a major stablecoin's failure transmits stress to dozens of derivatives that looked safe in isolation.
Pharos tracks 268 core stablecoins and cash equivalents, plus 97 variants, 27 stable-value investments, 44 upcoming launches, and 123 dead ones, then scores the core universe with honest governance classification, transitive dependency scoring, and live reserve composition where available. Real-time depeg detection, freeze monitoring across 34 stablecoins, and a 30-minute ecosystem-wide stability index surface stress as it builds, before it reaches the headlines.
When a tracked stablecoin effectively dies (issuer abandonment, supply trending to zero, irrecoverable depeg, regulatory shutdown), Pharos freezes it rather than deletes it, and surfaces it in the cemetery with an obituary and an archived detail page.
Pharos is a public good. The dashboard stays free, the code is open source, and sustainability comes from community support plus future paid API access for heavy programmatic usage. See the funding ledger.
For asset eligibility and lifecycle decisions, read the listing policy. For methodology and API references, use the documentation archive.
Operating stance
Pharos has no token, takes no issuer money, sells no paid placement, and accepts no editorial trade. Sustainability comes from community support and, eventually, paid API access for heavy programmatic users. The code is MIT-licensed and the inbound funding ledger is published in full at /funding/.
Most trackers show what a stablecoin is worth right now. Pharos shows what could make it fail: backing, governance, freeze authority, dependency exposure, peg behavior, and exit liquidity. Price is the current outcome. Pharos measures the conditions that can change it.
Safety Scores apply dependency limits through V10 parent and backing relationships. Serial claims inherit the required parent's final score, while collateral baskets inherit upstream backing quality in proportion to their mapped shares. The dependency map shows direct collateral and wrapper links, not a failure simulator. The 2023 USDC depeg showed why those links matter.
Every score Pharos publishes carries a methodology version and a changelog link. When a formula changes, the version increments and the change is documented before the new number replaces the old one. The project changes its mind in public so that anyone who cited an earlier reading can verify what changed between then and now.
The dashboard, the worker pipeline, the scoring engines, and the data exports are MIT-licensed and reproducible. What is tracked is a curatorial decision, and the curation is published. Additions land with explicit rationale, exclusions are documented when contested, and the inbound funding that supports the work is logged at /funding/ for inspection. Editorial content, including profile prose, methodology and case studies, is licensed under CC BY 4.0. Third-party material retains its original terms.
Narrative summaries on coin pages and the daily digest are drafted by a large language model against the same data the dashboard renders. Each panel names the model and facts date; a reviewer is named only after approving that exact text. Static adoption, funding, address, TVL, supply, and market-cap claims require a displayed source, date, chain scope, and denominator. Address counts are not presented as users without a dedicated methodology, and project financing is not compared with token supply unless the quantities are economically comparable. The numeric outputs, including scores, peg deviations, supply, liquidity, and freezes, are computed by the worker pipeline, not by the model.
Pharos publishes data that other people will cite, which means it has to be wrong in public when it is wrong. If a number is off, a classification is wrong, or a methodology choice does not survive scrutiny, the correction path is the same as for every Pharos data point: a GitHub issue, the on-page feedback link, or a message via Telegram. Verified corrections are applied and dated. See the funding ledger for the inbound side of the same accountability.
The team
TokenBrice created Pharos and leads product direction, scoring methodology, and the data pipeline. TokenBrice also works on Polaris, and Polaris-related coverage, like everything else on Pharos, remains purely data-driven. Ike drives growth and communications. Most of the codebase is written and maintained by Claude and Codex, from the worker pipeline and scoring engine to the frontend and daily digest.
In the wild
Prefer to watch? Appearances marked “Full walkthrough” are unscripted sessions of someone driving the dashboard live, hosted elsewhere.
Featured
Ethereum Builders Live: Pharos
Ethereum Foundation ·
The Ethereum Foundation hosted TokenBrice and chaskin.eth to explore what makes a stablecoin trustless, walk through the Pharos risk dashboard, and dig into stablecoin analytics.
Watch the broadcast
TokenBrice - Polaris & Pharos
Stable School (Stable Summit) ·
TokenBrice joined Stable School for episode 48 to discuss Polaris and Pharos.
Don't Touch A Stablecoin Without Checking Pharos First
DeFi Dad · · Full walkthrough, 69 min
A DeFi Frontier episode on using Pharos to vet a stablecoin before putting money behind it.
How Stable Is Your Stablecoin?
Leviathan News · · Full walkthrough, 88 min
TokenBrice walked through Pharos live — the motivation behind the project, the data pipeline, and how the main risk signals should be read in practice.
Coverage
Signals
Telegram alerts cover DEWS state changes, depeg events, and safety grade changes.
Open @PharosWatchBotImmersive data visualization
Governance lens
Pharos classifies stablecoins into three governance tiers: CeFi (fully centralized), CeFi-Dependent (decentralized infrastructure but reliant on centralized collateral or peg mechanisms), and DeFi (fully on-chain, no centralized custody dependency). The classification reflects actual infrastructure dependency, not marketing claims.
Editorial designation
Pharos Bluechip is an editorial roster built from external Bluechip reference ratings intersected with Pharos report-card grades. A coin qualifies only when Bluechip's synced rating is A-tier and Pharos independently rates the same asset A-tier or better.
The Bluechip grade itself comes from the external Bluechip data sync. Pharos does not recompute that methodology or expose a hidden Bluechip score; it stores the synced grade, date, and reference metadata as a separate input.
The roster adds a Pharos guardrail on top of that input. If an asset is not A-tier in Pharos report cards, it stays off the roster even when the external Bluechip feed still carries a high grade. That keeps Pharos from presenting an external rating as if it were a Pharos-owned floor result.
Each gate is independent. A coin clears Bluechip only when every gate clears at the same moment, and only while it continues to do so.
External Bluechip tier. The synced Bluechip reference rating must be A-tier. Pharos treats that grade as source data, not as a locally recomputed methodology.
Pharos report-card tier. The same asset must also hold an A-tier Pharos V10 grade (A-, A, or A+). This independent filter reflects Backing, Exit, and Economic Control, constrained by peg behavior, dependency exposure, evidence quality, and structural caps.
Tracked asset match. The Bluechip entry must resolve to a Pharos-tracked stablecoin ID. Unmatched feed rows stay out of the public roster until the mapping is explicit.
Fresh reference data. The roster is refreshed from the Bluechip sync and Pharos report-card API. Stale or unavailable inputs fail closed rather than inventing a substitute grade.
No local score blend. Pharos does not average the external Bluechip grade with its own score. The roster is an intersection: both systems must currently clear the A-tier threshold.
Roster suspended while the V10 grade floor is under review. It resumes under the same rule: the live intersection of external Bluechip A-tier ratings and Pharos A-tier report cards, current-state rather than historical. Bluechip rides the Safety Score methodology (v10.40) and is re-derived, never grandfathered, when a floor changes. The per-coin badge still links to the external Bluechip report.
Source flow
All data is fetched server-side by a Cloudflare Worker and cached in D1. The browser never calls external APIs directly.
Source groups
Aggregators, exchanges, oracles, NAV publishers, and direct DEX quotes.
Hive internal market (HBD/HIVE), with a CoinGecko HIVE/USD conversion dependency: one soft CoinGecko-family observation, not two independent votes or a conversion-price guarantee. DefiLlama, CoinGecko, GeckoTerminal, CoinMarketCap, DexScreener, DexPaprika, Alchemy Prices API, Moralis Token Prices, Birdeye, Jupiter Price API, Binance, Kraken, Bitstamp, Coinbase, Kava Pricefeed, Curve on-chain, Chainlink NAV reserve telemetry, Superstate NAV/liquidity telemetry, JPMorgan exact-class money-market NAV telemetry (JLTXX remains quarantined; staged bootstrap evidence is not price/market-cap admission), Flying Tulip's on-chain ftUSD reserve index, Fluid, Balancer, Curve, Uniswap V3 (including guarded USDU/USDT QuoterV2 sell quotes), Uniswap V4, Raydium, Orca, Meteora, PancakeSwap, Aerodrome, Velodrome, Aero, guarded Mento FPMM and Broker quotes, exact Aero BD/USDC sell quotes, reviewed Jupiter USDv/USDC direct-route quotes with independent Solana pool-state checks, dEURO EURC StablecoinBridge redemption, Citrea StablecoinBridge, Zephyr Scanner, Movement REST, direct protocol redemption or FX-par quotes, curated fail-closed on-chain supply-gap repairs, and V10 lockbox attribution reads. Exchange spot prices remain. CoinGecko discovery tickers supply price/observed-flow evidence only, weighted by flow and confidence with zero liquidity TVL; synthetic orderbook pools no longer count. Direct Binance/Coinbase/Kraken depth diagnostics and the dormant exact Kraken observer are retired, not scoring or settlement evidence
Issuer and protocol reserve APIs, attestations, proof-of-reserve feeds, and on-chain vault reads. Direct Morpho lender vaults add a finalized, hash-pinned, exactly reconciled NAV-accounted collateral-book census every four hours; unavailable books retain a typed reason, not an assumed safe allocation. Midas mTBILL joins an issuer SAVE registry checkpoint with its exact historical portfolio; signed date and aggregate do not independently assure composition, custody or liabilities. Axis's issuer asset-class transparency publishes five reviewed issuer classes plus, when the issuer reports an unclassified liquid-hedge amount, one residual row rather than venue inventory; MegaUSD custody is a reviewed custodian roster read at one pinned block, with the issuer JSON used only to detect roster drift.
Issuer and protocol reserve APIs, dashboards, proof-of-reserve portals, issuer attestation indexes, and direct on-chain vault/accounting reads (including Ethena collateralization and proof-of-reserves measurements, plus reserve composition disclosures and scoped feeds from providers such as 3Jane, Anzen, Chronicle Proof of Asset, Falcon, Frankencoin, Hashnote, infiniFi, M0, Makina strategy/allocation APIs with identity-verified on-chain Machine AUM reconciliation, Mento Reserve / analytics API, OpenEden cUSDO (TBILL's separate configuration is retained but quarantined, not an active feed), Origin OUSD vault/redemption, Blast, Nest Credit (positions and corroborated gross/net NAV), Re, Resupply, Reserve Protocol, USDD, USD.AI, USD1 Chainlink bundle oracle, Backed (public assetReserves circulation feed), Accountable, Axis issuer asset-class transparency (five issuer classes plus a disclosed unclassified liquid-hedge residual; class labels do not establish venue identity, asset-level custody or redemption), MegaUSD on-chain custodian-roster balance reads at a pinned block (the geo-gated issuer JSON is roster discovery only and never publishes an amount), LedgerLens / Wyoming FRNT, Coinbase ONED proof-of-reserves GraphQL, BlackRock BRSRV holdings CSV, ChinaAMC CUMIU curated factsheet composition (its disclosed pooled-fund holdings JSON is not ingested: duplicate holding row, 103.19% total and missing fund NAV), Universal USDU, Supercoin ZARSC, Secured Finance USDFC, Somnia BrandedCustodian, Hyperbeat, Tether, Frax, Circle, First Digital Labs, Ripple/Deloitte reserve examinations, Ondo OUSG and USDY dated portfolio disclosures, Midas mTBILL position disclosures, J.P. Morgan Asset Management's exact Token Class NAV and class-assets publisher (issuer-reported valuation, not independent portfolio assurance; JLTXX stays quarantined), Macropod issuer-signed reserve verification (issuer self-verification, not independent assurance), Solomon Labs token-backing API (selected Chancery reserve accounts for replacement USDv only; issuer-reported, not live RPC, a whole book or supply/redemption coverage), SG-FORGE, Paxos/KPMG reserve examinations, Money on Chain DOC accounting reads, Aura Partners, KPMG/SALVUS (Schuman Financial), Sky/MakerDAO, Chainlink PoR/NAV oracles, StraitsX / KK Yap & Associates, Kinesis, Quantoz, Yamato, Aave GHO, BIMA reserve telemetry, SMARDEX, f(x), Asymmetry, JupUSD, Morpho vault liquidity, USDGO, Yield Optimizer, Yuzu, Solstice, River, Alloy, Zephyr Scanner, Spiko, United Stables, Polymarket PUSD vault reads, Gnosis xDAI bridge collateral reads, Hive consensus-state reads, LeverUp LVUSD Monad vault census reads, Blox MYRC hash-pinned independent examination reports, Spark V1 local receipt inventory, Forest Road manager accounting reconciliation, and Curve/Yield Basis reserve reads where available; receipt/accounting measurements do not by themselves establish executable redemption liquidity). Single-asset v2 reports native CAD/QCAD and GBP/tGBP quantities on reviewed nominal bases, never USD; an Ethereum supply clock is diagnostic, not whole-reserve freshness, and reserve snapshots do not invent redemption terms or capacity. The six raw Chainlink PoR v3 bindings are BC3M, BIB01, GLDY, KAG, KAU and TUSD, preserving exact deployment inventories, original feed/supply clocks and typed unavailable reasons; BC3M uses a reviewed inclusive 86,400-second reserve/supply skew allowance after complete exact circulation reconciliation, while BIB01's unreviewed temporal policy withholds ratios. Separate Chainlink NAV sources are not part of that six-binding inventory. ZARP, Ripio wFIAT, USDU and ZARSC report indexes validate report identity/date while allocations remain reviewed static composition, not observed redemption. AUDD's agreed-upon-procedures report provides no assurance opinion; BRLV/AUDM are issuer-attested. Zephyr, AFI, ONED and BRSRV are limited issuer/protocol/disclosed-securities telemetry, not independently complete legal reserve books; static/weak sources remain detail/status evidence outside independent scoring admission. Solstice totals/proof strings are not a verifier or assurance stream; River's protocol-TVL quotient is not satUSD backing (genuine exit routes remain). DGLD adverse arithmetic stays degraded; XAGm retains original issuer-round age and its 60-day cap. ONRE's three reconciliation guards alone do not establish independent book evidence; Avant gross contextual observations/nulls retain wholly unknown composition. Flying Tulip reads one first-party index with three payloads, not independent chain inventory; SODAX's local census omits ICON/Archway/Havah/Injective and supplies no global ratio; FRNT gross supplemental metrics are not a retrieved assurance report or net-payable allocation. MYRC's independent examination preserves native MYR cash/fund rows, net-liability reconciliation and the original balance date under a 33-day cap; signing/upload/access does not renew balances. USDY reads the dated issuer portfolio categories with original-date freshness; SHVon remains tokenized-security exposure and other cash/equivalents stays unresolved. This composition does not reconcile all USDY liabilities or establish financial assurance or redemption capacity; LLC reports remain separately reviewed with manual refresh. Matrixdock STBT's reserve adapter, Origin's manual observer, Solomon legacy, USDH, Abracadabra MIM, StoneYield and OpenEden USDO runtime tails are retired; coin/curated/history evidence and replacement bindings remain
Block explorers, RPC providers, and pinned contract reads for supply, freezes, and flows.
Etherscan v2 (freeze events), TronGrid (JSON-RPC fallback for Tron freeze and USDT mint/burn events, plus the freeze event API fallback), Alchemy, dRPC, Dwellir (first eligible scheduled provider for supported EVM method/pin requirements and native operations, with independent diagnostic parity, key/budget/circuit admission and chain rollback; incumbent fallback order and inventory boundaries retained; Tron JSON-RPC freeze and USDT Issue/Redeem scans stop at the solidified head; mint/burn publication waits for 90-day admission; BSC/Monad/zkSync recent state does not establish archive history), selected public chain RPCs (including MegaETH public RPC, Fraxtal public RPC for reviewed supply and lockbox attribution without changing aggregate supply, EVM RPCs for configured mint/burn flows, direct Liquity/B.Protocol branch debt reads, and Frankencoin's ZCHF -> CHFAU StablecoinBridge balance probe, plus Solana mainnet RPC reads for tracked mint-supply validation, Starknet RPC reads, Sui mainnet GraphQL reads of Ember eEARN native receipt supply and USDY coin supply, Aptos REST reads of USDY legacy CoinInfo at a pinned ledger version, Stellar Horizon asset statistics (trustlines, claimable balances, liquidity pools and Stellar Asset Contract balances), read-only Soroban simulations of Spiko native share total_supply via the public Gateway RPC, Noble bank supply via Polkachu LCD (including IBC escrow, without adding remote vouchers), and DFINITY ICRC REST indexer reads for ICP), and reconciled freeze-ledger bootstrap rows from kyc.rip / stables.rip for major ETH and TRON blacklist coverage; Astherus asUSDF's public BSC (BNB Chain) EVM RPC with a Multicall3 aggregate3 read of the custom asUSDFEarn contract (not ERC-4626) to observe USDF backing net of unvested yield; Initia interwoven-1 LCD REST reads at /cosmos/bank/v1beta1/supply/by_denom, /initia/move/v1/view/json, and /initia/move/v1/accounts/{address}/resources/by_struct_tag to observe iUSD's AUSD0 vault backing and pin vault/metadata identities (Pharos's first Initia read path); and pinned direct route-controller/queue reads for Forest Road USDfr, Apyx apyUSD, Lido earnUSD, Monetrix USDM and Saturn sUSDat, with diagnostic versus measured capacity kept separate from reserve composition
External ratings, chain-risk snapshots, FX rates, commodity prices, and benchmark rates.
Bluechip, eurostablecoins.xyz EUR stablecoin coverage, VNX legacy vnx.li exchange suspension and current VNXAU issuer/service notices (exact-channel evidence, not a suspension of all holder exits), L2BEAT static chain-risk and Interop snapshots for Chain Health and reviewed Safety Score bridge-route context, Chainlink Data Feeds, ECB via Frankfurter, Open Exchange Rates (real-time FX cross-validation), fawazahmed0/currency-api (CNH and non-ECB FX), ExchangeRate-API (tertiary full-set FX fallback), gold-api.com, FRED DGS3MO, New York Fed EFFR, FRED DFF fallback, Treasury.gov yield curve XML fallback, the ECB Data API for 3M compounded €STR, SIX delayed SARON compound-rate downloads via public guest access, FRED and ALFRED IUDZOS2 SONIA Compounded Index mirrors with Bank of England IADB IUDZOS2 fallback (GBP SONIA Compounded Index), Bank of Japan Time-Series Data Search STRDCLUCON (JPY call-rate proxy), Banxico SIE SF43936 (MXN CETES 28d, token-gated), BCB SGS series 11 with sequential series 1178 fallback (same effective BRL SELIC, daily compounded / already annualized on a 252-business-day basis), Reserve Bank of Australia F1 money-market CSV (AUD cash-rate target), Bank of Canada Valet V122530 (CAD CORRA proxy), Central Bank of Russia DailyInfo KeyRateXML (RUB key rate), and CBRT EVDS BIST TLREF TP.BISTTLREF.ORAN (TRY overnight reference rate)
EU MiCA and national authority registers, plus U.S. GENIUS Act implementation sources.
ESMA MiCA registers, EBA EMT/ART issuer and significant-token registers, national competent authority registers such as ACPR REGAFI, DNB/AFM, BaFin, MFSA, CBI, and the Bank of Lithuania where relevant, plus U.S. GENIUS Act implementation sources such as OCC bulletins, FDIC rulemaking notices, FinCEN/OFAC AML rulemaking materials, Treasury state-regime comparability materials, OCC charter/decision materials, Federal Register notices, and issuer reserve or disclosure pages
Yield and liquidity APIs, subgraphs, and on-chain pool reads across DEXs and lending markets.
DeFiLlama Yields & Protocols (including exact Ethereum Uniswap V4 pool identity joins), protocol-native yield APIs and deterministic on-chain yield readers (Hashnote, Ondo, Midas NAV oracles, Re Protocol, Morpho, Pendle, Royco Dawn, Yearn Kong, Beefy, Aave V3, Compound V3, Curve scrvUSD current-rate, B.Protocol LQTY-only, Zephyr Scanner), Curve Finance API, The Graph (including live BSC recovery census and permanent execution-only input, not fee/price coverage), Fluid API + DexReservesResolver, Balancer API, Raydium API, Orca API, Jupiter direct-route quotes, Meteora API, Bluefin ordinary liquidity/discovery, Stellar Horizon classic-AMM pools, Aquarius's eight reviewed Spiko Soroban identities, TzKT Tezos uUSD holder/reserve census, Balanced bnUSD pools on ICON, Kava x/swap native USDX pools, PancakeSwap subgraphs, SUN.io SunSwap V2 pool census plus Smart Router and pinned V2 Router proofs with TronGrid RPC state, reviewed Uniswap V3, PancakeSwap V3, and Aerodrome Slipstream QuoterV2/factory RPC reads, Aerodrome and Velodrome Sugar view contracts, Aero factory and Metaquoter RPC reads, ordinary current/legacy Shadow Exchange venue data, GeckoTerminal, DexScreener. Listing a venue or discovery source does not activate measured routes: Ethereum alone scores hook-free V4, BSC/Base/Arbitrum/Polygon/Tempo collect shadow evidence, Unichain measurement is retired, and hooks remain unsupported negative identity/collision evidence. Hybra/XSwap exact V3, Sonic legacy and Base/Optimism volatile exact Solidly diagnostics are retired; Base/Optimism stable diagnostics remain non-scoring, without broadening classic-stable or Slipstream scoring coverage. DLMM repeated collection and Bluefin scheduled Sui selection stop, but ordinary venue sources remain; Orca/Raydium CLMM and Cetus diagnostics remain non-scoring and Raydium Standard/CPMM provider diagnostics stay shadow. Curve composite/NG preparation activates no cohort and retires none of the six retained observers; modeled DOLA/OUSD routes remain modeled and any future approved sell direction must be named exactly. vaults.fyi and dormant BIMA Earn/Etherfuse CETES yield collectors are retired, not current optional sources; Banxico's independent MXN CETES benchmark remains. Dead or deprecated DEX slugs such as Bunni are blocked from runtime pricing and liquidity inputs rather than treated as live venues
Processing path
Sources
Market, on-chain, ratings, FX, commodity, and digest inputs are collected on a fixed schedule.
Cloudflare Worker + D1
Staggered 5-minute, 15-minute, 30-minute, hourly, multi-hour, daily, and monthly lanes normalize the data, preserve independent protocol evidence in DEX price challenges before bounded TVL coverage selection, and cache the results for the public API.
Static dashboard
Next.js pages on Cloudflare Pages consume the worker outputs and render the stablecoin view without direct third-party calls.
Scoring details
Pharos Safety Score V10 grades stablecoins across Backing Quality, Exit Strength, and Economic Control, then applies peg behavior, deployment adjustments, and binding caps. The methodology page covers the full grading formula, peg score computation, DEX liquidity scoring, and dependency propagation.
Read the full methodologyPharos is a project by TokenBrice, Ike, Claude, and Codex. It puts the stablecoin data you want to monitor in one place: honest classification, freeze tracking, and a graveyard for the ones that didn't make it.
Pharos uses 268 core stablecoins and cash equivalents for ecosystem market aggregates, while keeping tracked variants and stable-value investments visible as separate listing classes. It documents 123 dead stablecoins in the cemetery, monitors issuer freeze and blacklist events on-chain, provides peg scores and depeg heatmaps, scores DEX liquidity, computes a 30-minute Pharos Stability Index, and publishes risk report cards.
Pharos classifies stablecoins into three governance tiers: CeFi (fully centralized), CeFi-Dependent (decentralized infrastructure but reliant on centralized collateral or peg mechanisms), and DeFi (fully on-chain, no centralized custody dependency). This reflects actual infrastructure dependency, not marketing claims.
Pharos aggregates data from DefiLlama, CoinGecko, on-chain RPC nodes, Etherscan, TronGrid, protocol-native APIs, public regulatory registers, and curated sources like Bluechip. Details on all data sources are available on the About page.
Reach out
Pharos code and data are MIT-licensed. If you spot a bad data point, want a stablecoin added, or want to understand how something is computed, open the code or reach out directly.
Press and research enquiries: [email protected]
