Language reference

Describe the strategy.
Driver trades it.

DexScript is a small, Pine-inspired language for writing a Solana trading strategy as a set of rules. You don’t pick a token; you describe the behavior, and Driver continuously scouts the best-matching token and trades it within your risk limits. Write it by hand, or describe it in plain English and let the AI draft it.

Generate from a promptWrite it by hand

Anatomy of a strategy

Every strategy is one block with a few labelled sections.

sol-trend-rider.dex
// A complete strategystrategy "SOL Trend Rider" {  universe:  auto            // best market match, scouted continuously  timeframe: 15m   buy:    when rsi(14) < 32 and crossover(ema(9), ema(21))    size: 20% of allocation   sell:    when rsi(14) > 70    stop_loss: 8%}
strategy "name" { }
The wrapper. Required; the name is any quoted string.
universe: auto
What to trade. auto lets Driver scout the best match continuously.
timeframe: 15m
The candle interval rules run on, like 5m, 15m or 1h.
buy:
When to open (buy) a spot position. Required.
sell:
When to close (sell) it. Required.
when …
The condition that triggers a buy or sell.
size: 20% of allocation
How much of your funds to commit per position.
stop_loss: 8%
Risk exit if the trade moves against you.

The vocabulary

Keywords & values

strategyuniversetimeframebuysellentryexitwhenandorsizeofallocationstop_losstrailing_stopprice
autotruefalse

Combine conditions with and / or, compare with < > =. Numbers, percentages and quoted strings are literals; // starts a comment.

Indicator functions

The vocabulary you build conditions from.

  • rsi(14)Relative Strength Index. Overbought / oversold.
  • ema(9) · sma(20)Exponential / simple moving averages.
  • macd()MACD line vs signal. Momentum.
  • bbands(20, 2)Bollinger Bands.
  • atr(14)Average True Range. Volatility.
  • volume() · avgVolume(20)Current volume and its average.
  • crossover(a, b)True when a crosses above b.
  • crossunder(a, b)True when a crosses below b.
  • stoch()Stochastic oscillator.
  • funding()Perp funding rate.
  • forecast(24)Predicted price 24 bars ahead (TimesFM).

Common patterns

The four signal families Driver runs. Each token is scored 0 to 100%, and bought at 100%.

RSI reversion

Buy dips, sell rips
buy: when rsi(14) < 30sell:         when rsi(14) > 70

EMA momentum

Ride trends
buy: when crossover(ema(9), ema(21))sell:         when crossunder(ema(9), ema(21))

MACD momentum

Trade the cross
buy: when crossover(macd(), 0)sell:         when crossunder(macd(), 0)

Breakout

Buy strength
buy: when price > bbands(20, 2)              and volume() > avgVolume(20)sell:         when crossunder(price, ema(20))

Forecasting with TimesFM

Beyond reacting to indicators, Driver can act on a prediction. forecast(n) returns the token's expected price n bars ahead from Google's TimesFM, a 200M-parameter time-series model, so a rule can trade what the model sees coming.

sol-forecast.dex
// Trade the model's forecaststrategy "SOL Forecast" {  universe:  auto  timeframe: 1h   buy:    when forecast(24) > price * 1.03   // model sees +3% over 24h    size: 20% of allocation   sell:    when forecast(12) < price          // forecast turns down    stop_loss: 8%}

Zero-shot, numeric

TimesFM forecasts the price series directly. It is a dedicated time-series model, not an LLM, so predictions stay quantitative and reproducible.

Point and range

forecast(n) is the point estimate; the model also yields P10–P90 quantiles Driver uses to gauge confidence before sizing a trade.

Composable

Mix it with indicators: gate a breakout on a positive forecast, or tighten exits when the forecast rolls over.

Forecasting runs as an opt-in agent skill (off by default). It needs a TimesFM inference service; until that is wired, forecast() validates and highlights but returns no live prediction.

Generate from a prompt

Describe the strategy in a sentence or two and the AI drafts the DexScript, which you can then edit. Good prompts name four things: the edge, the entry trigger, the exit, and the risk appetite. The risk level you pick also tunes sizing and stops.

  • “Buy oversold pullbacks within an uptrend, take profit quickly, cut losers fast.”
  • “Momentum: enter when the 9 EMA crosses above the 21 EMA on 15m, exit on the reverse cross, aggressive sizing.”
  • “Mean-reversion on high-volume memecoins: buy when RSI is under 30 with above-average volume, sell at RSI 65, tight 5% stop.”

Write it by hand

Pick Write DexScript in the create wizard for a starter template, syntax highlighting, and a live function reference. The editor validates as you type and blocks publishing until the script is valid.

A valid script needs

  • A strategy "name" { … } block
  • An entry rule
  • An exit rule
  • Balanced { } and ( )