Dunamu launches AI-powered backtesting tool for crypto trading strategies

A new service lets users test trading strategies through conversation with AI.

Dunamu, the operator of digital asset exchange Upbit, announced Tuesday the beta launch of the Upbit Strategy Toolkit, an AI-based backtesting tool.

The service lets users enter investment ideas in plain, conversational language and verify them against actual historical market price data. It is designed to help users build and test hypothetical investment strategies.

A notable feature is that the toolkit works in conjunction with external AI services users already subscribe to individually. Users can simulate a strategy's returns, maximum drawdown (MDD) and other metrics based on historical price data, allowing them to assess a strategy's hypothetical performance before committing any capital.

The Upbit Strategy Toolkit separates the roles of the AI and the backtesting engine — a design intended to reduce "AI hallucination" problems, in which an AI fabricates price data or applies inaccurate calculations to produce distorted results.

The AI translates a user's natural-language trading idea into a set of calculable strategy rules, while the toolkit's built-in backtesting engine handles the actual performance calculations using historical price data. Dunamu said the tool delivers backtesting results that are consistently reproducible under identical conditions, rather than arbitrarily generated outputs.

The toolkit also supports 15 major technical indicators, including the Relative Strength Index (RSI), Moving Average Convergence Divergence (MACD) and Bollinger Bands, and can analyze data across a wide range of intervals — from one-second candlesticks to monthly candles. Factoring in per-market fees and trade execution conditions, it delivers key performance metrics such as returns, MDD and win rate in the form of a research note.

Users can compare multiple strategies by adjusting parameters through conversation with an AI agent. For example, after entering an idea such as "buy when RSI falls below 30 and sell when it rises above 70," users can modify the time period or indicator thresholds and observe how the results change.

In addition, the Upbit Strategy Toolkit supports integration with AI coding agent environments that users operate individually, including Claude Code, Cursor and Codex.

"A key strength is that we have significantly lowered the barrier to running backtests with AI while substantially improving the reproducibility of results," an Upbit official said, adding that the company will continue to pursue technological innovation to deliver more advanced systems and a more efficient user experience.

The Upbit Strategy Toolkit is a backtesting tool and does not recommend specific investment strategies or guarantee returns.

Meanwhile, Dunamu has been expanding its use of AI technology, including the full rollout earlier this year of an "AI investment" service in which AI understands an investor's intent and autonomously buys and sells digital assets.