Infrastructure · Strategy Backtester
Backtest any strategy by describing it.
An MCP agent runs cross-asset backtests from a plain-language description, over options, futures, perpetuals and spot, priced on theoretical mids and on real listed bid/ask.
- A single prompt over MCP
- Theoretical mid and real listed bid/ask
- Hedge attribution, Sortino, max drawdown
- Deterministic — identical queries, identical outputs
Watch the agent work
Several days of data, analysed in minutes.
Ask in plain English. The agent retrieves the surface, compares it over time and returns the signal, a task that would otherwise take an afternoon.
Has the BTC 25Δ risk reversal flipped? Compare to a week ago.
Yes — the 1-week 25Δ risk reversal flipped to +1.8 vol (from −0.4 a week ago): calls are bid over puts for the first time in 3 weeks. Front-end term structure steepened ~4 vol — skew is turning call-heavy.
Watch the agent work
Several days of data, analysed in minutes.
/01
Ask in plain English
Describe the trade in ordinary language, without a query language or schema.
/02
It pulls days of data
90 days of the BTC IV surface, 42,318 quotes, fetched live via MCP.
/03
Compares the surface
25Δ risk reversal and term structure, now versus a week ago.
/04
Surfaces the signal
Risk reversal moved to +1.8 vol, with calls bid over puts. Completed in 2m 14s.
Has the BTC 25Δ risk reversal flipped? Compare to a week ago.
Yes — the 1-week 25Δ risk reversal flipped to +1.8 vol (from −0.4 a week ago): calls are bid over puts for the first time in 3 weeks. Front-end term structure steepened ~4 vol — skew is turning call-heavy.
What you get
Every backtest reads from the same institutional data engine.
Pricing
Dual mode
Theoretical model mid and real listed bid/ask, so the same strategy can be run both ways.Assets
Cross-asset + RWA
Options, futures, perps and spot — plus RWA equity and commodity feeds, and FX perpetuals.Venues
Exchange-sourced
Deribit, OKX and Bybit are among the connected sources combined into a composite.Drive
Natural language
A single prompt over MCP.Metrics
Full ledger
Trade-by-trade P&L, hedge attribution and Sortino, from the fills themselves.Determinism
Reproducible
Identical queries return identical outputs, so a result can be re-run and audited.
Pricing
Two pricing modes.
Illustrative straddle · premium quoted in BTC
Illustrative equity curves for demonstration: the same strategy filled at a theoretical model mid returns +31.4%, and filled at listed bid/ask returns +24.6% — a spread drag of −6.8%. The listed curve is constructed as a fixed fraction of the model-mid curve, so it never sits above it.
+31.4%
Model mid
+24.6%
Listed fills
−6.8%
Spread drag
Runs on
A backtest is only as reliable as the record underneath it.
- /01
The record it tests against
Point-in-time, no lookahead, with delisted instruments retained so that failures remain in the record.
- /02
How far back it reaches
Point-in-time history to 2020, on the same schema the live endpoints serve.
- /03
How you drive it
The same Model Context Protocol server that serves prices, surfaces and indices to an agent.
Use cases
What teams build on the Strategy Backtester
- AI-Driven Strategy Backtesting via MCPAI agents design, execute, and iterate on cross-asset derivatives backtests through MCP — mixing options, futures, perpetuals, and spot in a single conversation. Theoretical and listed pricing modes reveal the gap between model edge and executable edge.Read the story
- Prop TradingBacktesting Systematic Options Strategies with Historical Vol DataBacktesting systematic options strategies — delta-hedged straddles, calendar spreads, vol targeting — across years of calibrated historical volatility data. Test strategy performance across multiple market regimes before deploying capital.Read the story
- Hedge FundSystematic Strategies: Identifying Vol Regime Shifts in Crypto DerivativesSystematic strategies that identify volatility regime shifts through term structure spreads and skew metrics across crypto derivatives markets. Evaluating the spread between 30-day and 7-day implied vol alongside 25-delta risk reversals to generate alpha.Read the story
Scale
150+
Tokens covered
250M+
Data points / day
1.8M+
Instrument listings catalogued
2020
History since
Get started
Describe a strategy and test it against the listed spread.