Not for traders who want signals. For traders who already build theses and want an AI that reasons with them.

Every answer includes your positions, your theses, and your past behavior — not just the chart.

Five layers of context stream into the model the moment you ask. No copy-paste. No screenshots. No translation layer.

01
Market state

Not just price

  • Current regime (1 of 12) with confidence + secondary regime
  • Basis (APR), funding, 25Δ skew, IV term structure
  • GEX positioning + flip level
  • IV vs RV premium — are options overpriced or underpriced?
“Is this a trend continuation or pre-event tension?”
A regime classification with the exact variables driving it — and what would invalidate it.
02
Your book

Actual exposure, not guesses

  • Spot + perps (long and short) in one view
  • P&L, R-multiple, risk flags
  • Distance from entry to invalidation
  • Scanner verdict on each open position
“Where am I leaning too hard?”
Exposure vs regime and derivatives positioning — e.g. “3 longs, all in negative-skew regime with rising funding.”
03
Your theses

Not disconnected trades

  • Thesis, direction, assets
  • Conditions engine — how many are actually met right now
  • Invalidation criteria (set at creation, not hindsight)
“Which of my plays are actually valid right now?”
“2 of 3 plays aligned. Your ETH long is fighting rising skew — 1 condition reversed since entry.”
04
Your behavior

The part most traders ignore

  • Journal entries with entry-time market context
  • Mistake tags: fomo, early_entry, chasing, oversized…
  • Performance by regime, setup type, and asset
“Am I repeating the same mistake again?”
Pattern recognition tied to real trades — e.g. “last 6 losses came from early entries in Vol Compression.”
05
Historical precedent

Not vibes

  • Conditional forward returns
  • Lead-lag relationships (e.g. DXY → BTC)
  • Top-5 historical analogues to current conditions
“When we looked like this before, what happened next?”
A distribution of outcomes — not a single prediction.
You’re not asking an AI for opinions. You’re querying your own system — with the full market and your full history attached.

Carrac works without AI too. Here’s what sits under the MCP layer.

Three moats that every serious trader benefits from — with or without a model in the loop.

I
Intelligence

Derivatives Regime Engine

  • 12 deterministic regimes from live Deribit data
  • Auditable fingerprints — no ML black box
  • Quant DB with 45+ metrics and conditional forward returns
  • Formula engine — cross-domain expressions in one line
87% in Healthy Trend. 41% in Melt-Up.
II
Workflow

Discover → Frame → Execute → Learn

  • Six-criterion scanner with explainable scores
  • Plays: thesis + conditions + invalidation, auto-evaluated
  • Unified spot + perp positions with live scanner verdict
  • Auto-populated journal with market context at entry
You don’t lose because the market changed. You lose because you didn’t notice in time.
III
Integrations

Bring your own AI

  • MCP server with 33 tools
  • Claude Desktop, Gemini, Cursor — same tools
  • In-app copilot if you don’t bring your own
  • OpenAPI spec, webhooks, CSV/JSON export
Your trading system, inside your AI.
270+ ASSETS COVERED

Across crypto, US stocks, ETFs, forex, commodities.

Live data from Binance, Hyperliquid, Finnhub, FRED, Deribit — with a multi-venue fallback chain so the chart never goes blank. Every number is auditable, every source is named.

No black-box data. Every source named.

270+ curated assets across crypto, US stocks, ETFs, forex, commodities, and indices. Multi-venue fallback so the chart never goes blank.

Binance Hyperliquid Bybit MEXC Gate.io KuCoin Finnhub FRED Deribit Yahoo Finance

Every number is auditable. Every source is named. See all data sources →

256-bit TLS encryption
Free tier, no card required
Payments via Paddle — cancel anytime

Stop prompting in a vacuum.

Free tier is generous because we want you to feel the difference. Bring your own Claude via MCP and the AI credit cost disappears.