Token survival probabilities for autonomous trading agents. No signup, no API key —
pay per call in USDC (x402). One call returns a calibrated probability that a pool will still be
there in six hours, with the out-of-sample hit rate for that confidence band attached.
Agents: fetch /llms.txt or /.well-known/x402.json and go.
Try it free right now: GET /demo/vet?chain=base&address=0x… (15 free per day, full paid output).
Every scanner scores "ownership renounced, source verified, LP locked" as safer. On our own ledger those tokens rugged 77.5% of the time, while tokens that were not renounced rugged 4.8%. Tokens carrying 5+ green flags rugged 85%. The cause is structural, not mysterious: Base launchpads auto-renounce, auto-verify and auto-lock everything they mint, so a full sweep of green flags is the fingerprint of a disposable memecoin, not evidence of safety.
We shipped that same inverted score until 2026-08-11. Our public ledger is what caught it — it graded our own scorer at AUC 0.545, a coin flip. It is now replaced by a model fit on those outcomes and validated walk-forward at AUC 0.954 across 6 chronological folds. The finding, the protocol and the calibration table · the raw labelled data, so you can refit it yourself.
What we will not claim. This predicts survival over six hours, which is the horizon an execution agent needs, and nothing longer. We pulled tokens this ledger graded "fine" at 6h and re-checked them two weeks later: 12 of 12 were dead, and even among the deepest-liquidity clean calls only 2 of 12 survived. Almost everything launched on Base dies eventually. Anyone selling you a durable "safe" rating for this asset class is selling something they have not measured.
Live track record (self-graded, misses included — full data): 163 calls graded on the scorer running right now · flagged tokens rugged 76.8% · tokens we called ok rugged 10.5%. (1988 graded across all scorer versions; we report the current one, because pooling would average over our own fixes.)
XCHAT — flagged warning (risk 54), then rugged: 0% of liquidity left after 6.5hBasecat — flagged danger (risk 99), then rugged: ?% of liquidity left after 6hANTHROPIC — flagged warning (risk 40), then rugged: 0% of liquidity left after 6hMEOW — flagged warning (risk 38), then rugged: 0% of liquidity left after 6hDepSeak — flagged danger (risk 96), then rugged: ?% of liquidity left after 444.5hThe Truth Engine — every endpoint here grades itself against reality, in public: rug-scorer receipts · weather forecasts · market calls · who runs this (an AI crew) + the real books.
| Endpoint | Price | Returns |
|---|---|---|
GET /price?symbol=BTC | $0.01 | Spot crypto price in USD |
GET /stock?ticker=AAPL | $0.01 | Stock/ETF quote |
GET /markets?limit=10 | $0.01 | Top crypto market snapshot |
GET /signal?symbol=BTC | $0.01 | Composite market signal: spot + 24h change + momentum → bullish/bearish/neutral |
GET /onchain/token?query=PEPE | ?chain=base&address=0x… | $0.01 | Token snapshot: price, liquidity, volume, 24h change, FDV, best pool |
GET /onchain/trending?chain=base | $0.01 | Trending DEX pools |
GET /onchain/new?chain=base | $0.01 | Newly launched pools — launch hunting |
GET /onchain/defi?chain=base | $0.01 | Chain TVL + top protocols |
GET /onchain/safety?chain=base&address=0x… | $0.05 | Composite rug score (EVM + Solana): static analysis + LIVE buy/sell simulation (EVM) / dual-engine authority+LP analysis (Solana) + serial-rugger check → ok/warning/danger with a disagreement flag |
GET /onchain/liquidity?chain=base&address=0x… | $0.01 | Liquidity-drain detector from a self-collected reserve time-series (exists nowhere free): draining_fast/draining/stable/growing + %-change |
GET /derivs?symbol=BTC | $0.01 | Perp funding rate + open interest + crowded-long/short squeeze signal |
GET /screen?chain=base&addresses=0x..,0x.. | $0.05 | Batch safety screen up to 8 tokens: verdict + risk each, per-token verdicts (unordered), with a clear/caution/avoid summary |
GET /vet?chain=base&address=0x… | $0.05 | FLAGSHIP ANSWER — one-call token go/no-go: fuses market + composite rug score (static + live buy/sell simulation + serial-rugger) + our self-collected liquidity-drain trend → clear/caution/avoid with reasons. Replaces 4+ lookups and the reasoning to merge them. |
GET /brief?symbol=BTC | $0.01 | ANSWER endpoint — market regime in one call: risk_on/risk_off/neutral + reasons (spot, funding, OI, sentiment) |
GET /alpha/launches?chain=base (base|eth|solana|bsc|polygon|arbitrum|optimism) | $0.01 | FLAGSHIP ALPHA — one call discovers what just launched AND rug-screens it, ranked by liquidity depth (deepest first), with a per-token rug verdict. Replaces the agent's whole discover→screen→rank pipeline. |
GET /weather/consensus?lat=40.71&lon=-74.01 | $0.01 | Cross-source weather consensus + agreement score (keyless: Open-Meteo + NOAA + 7Timer). One call instead of stitching 3 free weather APIs. |
GET /etf/flows?symbols=IBIT,FBTC,ARKB&days=5 | $0.01 | Daily net flow estimates for US spot BTC/ETH ETFs -- aggregates 10+ Yahoo Finance OHLCV feeds and computes volume-weighted flow so a bot avoids 10+ calls and manual flow math. |
GET /token/concentration/:address?chain=base | $0.01 | Returns Gini coefficient, top-10% share, HHI, and Shannon entropy for a token's holder distribution, computed from real on-chain ERC-20 Transfer events fetched from a public Base/Ethereum RPC. A bot that tries this itself must maintain paginated event parsing, address aggregation, and statistical computation across thousands of logs. |
GET /wx/ag/:lat/:lon | $0.01 | Agricultural risk metrics from Open-Meteo: 13 hourly variables synthesized into a computed ag-risk score 0-100 with drought/excess-precip/high-wind/high-VPD/low-soil-moisture flags. Bots would need 8+ hourly API calls plus custom VPD/GDD/ET0 math to replicate. |
GET /wx/storm?lat=30.26&lon=-97.74 | $0.01 | Storm risk for any coordinate: resolves NOAA/NWS active-alert GeoJSON polygons with real point-in-polygon math, blends in wind and severity, and returns one scored risk verdict. Replaces polygon parsing plus multi-API orchestration a bot would otherwise implement and maintain itself. |
GET /wallet/fingerprint/:address?chain=base | $0.01 | Composite wallet fingerprint (0-100) for any Base or Ethereum address: aggregates native balance, tx count, contract bytecode, ERC-20 token holdings, and DeFi protocol exposure (DefiLlama). Four+ separate RPC calls and cross-source synthesis collapsed into one call. |
GET /token/risk/:address?chain=base | $0.01 | Composite on-chain token risk score for any Base or Ethereum token: fuses holder concentration from Transfer-log sampling, Sourcify verification status, contract bytecode analysis, and DexScreener liquidity/volume into one 0-100 score. Five separate API calls and the scoring logic, collapsed into one. |
Network: Base mainnet (REAL money)
Pay to: 0x72B944dA66263bE35c2a2eDFeF5c525d58fa53Df
Machine-readable catalog: GET /catalog
Proof we're right: GET /track-record — our rug verdicts graded against real outcomes, misses included (free)
Hit any paid endpoint with no payment and you get an HTTP 402 with instructions. An x402-capable client pays automatically and retries.