Top 10 atomic signals ranked by 7-day cross-sectional IC, gated to A/B health grades and non-anti-predictive names. Each carries trigger stats and a 7d walk-forward IS/OOS robustness block. Runs across 200+ digital assets — BTC, ETH, SOL, top-cap + long-tail coverage.

Daily taker buy/sell USD and buy/sell ratio per asset, split into spot and futures arrays, last N days. Raw tape for custom flow models. Coverage: pass any asset in the CRYPTYX 200+ tracked universe (BTC, ETH, SOL, top-cap + long-tail).

Arbitrary trigger definition — POST any {type: z_score|z_differential|composite, ...} predicate and get the full institutional evidence envelope back. 5-year backtest, Sortino, max drawdown, profit factor, walk-forward IS/OOS, regime context, rolling hit rates. Enterprise tier because every request runs a live backtest (no cache). The extensibility endpoint: anything expressible in Metric Slicer becomes a triggerable institutional-grade primitive. Backtests run over the CRYPTYX 200+ tracked digital assets — BTC, ETH, SOL, top-cap + long-tail coverage.

Curated institutional-grade trigger presets — mean_reversion_price, vol_expansion_alert, vol_regime_break, flow_inflection, treasury_manager_classic. Each returns the full evidence envelope: 5-year backtest with hit rate, Sortino, max drawdown, profit factor, rolling 30/90-day performance, regime-conditional stats, last 10 triggered events. Nightly-refreshed for top-50 assets; live evaluation for others.

Cross-window z-score differential trigger — the generalized crossover primitive. Pass two metric_ids (typically same factor at different windows, e.g. VOL_RV_7D vs VOL_RV_30D) and a threshold. Fires on the delta z(a) − z(b). Same institutional envelope. Applies across every factor family: TR, VOL, FLOW, FUT, OB, CORR. Backtests run over the CRYPTYX 200+ tracked digital assets — BTC, ETH, SOL, top-cap + long-tail coverage.

Single-metric z-score threshold trigger — pass any metric_id, operator (gt/lt/gte/lte/abs_gt), and threshold. Returns the same institutional envelope as the presets: 5-year backtest, Sortino, MDD, profit factor, walk-forward, regime context. Applies to any of 440+ metrics across 8 factor classes. Backtests run over the CRYPTYX 200+ tracked digital assets — BTC, ETH, SOL, top-cap + long-tail coverage.

One-shot agent grounding snapshot in 200-400ms — factor breadth, top/bottom composite assets with regime, recent signal aggregation, platform counts, and pipeline status. Call before any market reasoning task. Runs across 200+ digital assets — BTC, ETH, SOL, top-cap + long-tail coverage.

Single-call LLM grounding bundle — top 10 and bottom 5 composite-ranked assets, all signals triggered today, 30-day factor breadth across 8 classes, plus a ready-to-drop plaintext brief for an agent prompt. Runs across 200+ digital assets — BTC, ETH, SOL, top-cap + long-tail coverage.

CRYPTYX is the intelligence layer for autonomous trading agents — the same institutional-grade research stack that powers professional desks, priced per call in USDC. 200+ digital assets, 440+ metrics, 150+ signals across 8 factor classes, walk-forward validated, macro-regime aware. Agents don't just consume signals. They backtest any factor over 5+ years of history (Metric Slicer), fork and validate custom signal recipes (Signal Forge), scan the universe for live matches, decompose per-asset conviction across every factor class, and query the full stack in natural language. The loop: define your conviction thesis → validate it against history → scan for live matches → execute via OKX, Kraken, or Coinbase execution toolkits. CRYPTYX is the brain; exchanges are the rails. No accounts, no API keys, no rate cards.

Every canonical setup evaluation for a given day, filterable by state (fires / boundary / silent / first_ever), factor family, or asset. Includes z_scores per metric, distance_to_trigger, confidence. Distance is measured in sigma — positive = past threshold, negative = short of threshold. Boundary + first_ever states are the highest-signal reads. Applicable across the CRYPTYX 200+ asset coverage universe (BTC, ETH, SOL, top-cap + long-tail); backtested scope is the 33-asset curated slice.

First-ever fires — setups that fired for the FIRST TIME in prospective history. Rarest, highest-signal events in the canonical catalog. A setup with state = first_ever has never fired before since it was added. Historical backtest reflects behaviour on ADJACENT setups or partial configurations, not this exact combo — treat as structurally important context, not directional recommendations. Applicable across the CRYPTYX 200+ asset coverage universe (BTC, ETH, SOL, top-cap + long-tail); backtested scope is the 33-asset curated slice.

Canonical setup catalog list — 150 founder-approved multi-metric compositions across 8 factor families + cross-family fades. Filter by factor_family (TR/EFF/VOL/FLOW/FUT/OB/OPT/CORR/CROSS_FAMILY), tier (1-5), direction (long/short/neutral), asset (in applicable_assets), or durability_class (structural for 90d+ high-Sharpe, tactical for small-sample, hybrid for the rest). Each setup ships display_name, composition, primary_horizon, backtest snapshot pointer. The shipping surface — not a raw metric database. Applicable across the CRYPTYX 200+ asset coverage universe (BTC, ETH, SOL, top-cap + long-tail); backtested scope is the 33-asset curated slice.

Canonical harness tear-sheet — one call returns totals (fires / boundary / silent / first-ever), top-5 fires by confidence, first-ever prints (5-year rarity), boundary watchlist (near-fire ranked by distance-to-trigger), and silent fade setups. Every setup carries plain-English composition (via ref.metric_display_labels) so responses never leak raw metric_id strings. Priced cheap so polling patterns stay economically viable — every MCP skill instructs Claude to call this FIRST as the canonical orientation. Scope: apex (BTC/ETH/SOL default), all (33-asset curated universe), or asset:{SYMBOL} for single-asset lens.

Near-fire setups ranked by distance-to-trigger DESC (closest to firing first). Primary monitoring surface for autonomous agents — cheap to poll at 15-30 min intervals to detect when a queued setup is about to fire. Distance in sigma; 0.2 means one metric is 0.2σ from its threshold. Cheaper than polling /api/v1/harness/evaluations when you only care about near-fires. Applicable across the CRYPTYX 200+ asset coverage universe (BTC, ETH, SOL, top-cap + long-tail); backtested scope is the 33-asset curated slice.

Service health probe — build version, last pipeline run, current signal/composite/asset counts. Sub-second, no DB writes. The cheapest end-to-end pay-per-call check before issuing real queries.