TILLA · model-ready market data

Machine-learning features for crypto

Skip the data-engineering project. TILLA is a model-ready feature matrix — 1,817 value-scored features on one 1-minute clock, so you spend your time on models, not pipelines.

The hardest part of a crypto model is rarely the model — it is assembling clean, aligned features. TILLA hands you the features directly: 1,817 columns, each scored on its single-feature predictive power versus next-day BTC / SPX / EUR-USD returns, so the most valuable signals are front and center.

Features, not raw dumps

• Engineered technicals — EMA distances, RSI, MACD, Bollinger width/%b, Donchian position, returns over multiple horizons and volatility regimes, across crypto, equities and FX.
• Cross-asset context — betas and correlations to DXY, gold and the 10-year, plus BTC/ETH ratio dynamics.
• Positioning & flow — funding, open interest, long/short, taker flow, COT and per-fund ETF flows.
• Regime & sentiment — Fear & Greed, VIX term structure, liquidity and calendar encodings (time-of-day, day-of-week) that are free at every tier.

Train on one aligned row per minute

Every feature shares one timestamp, so building an X matrix is a single call — no joins, no leakage from misaligned timestamps. Use /v1/sync to pull your whole tier window into a local store, then keep it fresh incrementally.

import requests
r = requests.get("https://api.tillacrypto.com/v1/sync",
    headers={"Authorization": "Bearer tk_live_..."},
    params={"since": "1970-01-01"})   # first pull = whole window
rows = r.json()["data"]               # model-ready, already aligned

See the exact shape of a full aligned row (all 1,817 columns) with no key at /download/max.json, or connect it straight to your agent via the TILLA MCP server.

FAQ