# StratCraft StratCraft is an open-source, AI-powered quantitative research platform that lets anyone who can describe a trading idea in plain language turn it into a statistically validated, production-ready C++ strategy. The platform ships with 60+ ready-to-use strategy templates, from classic approaches like Turtle Trading, Dual Thrust, Bollinger Mean Reversion, Pairs Trading, and Fama-French to ML models including XGBoost Ranking, Random Forest, and Reinforcement Learning (DQN), plus 138 built-in technical indicators. An LLM-driven pipeline discovers and scores signals across four rounds of validation, and the Alpha Factory fuses up to 2,000 of them through 12 weighting methods such as Sharpe-ratio ranking, information coefficient screening, and correlation-adjusted combination, the same multi-factor statistical approach employed by top quantitative hedge funds. The fusion layer starts from linear blends and evolves into an ML Combinator that learns, from accumulated signal scores and forward returns, how to weight every signal on every bar. Models train in Python and export to ONNX, so the C++23 engine runs inference with zero Python on the hot path. A streaming architecture processes 83 million execution-bar cells at a constant 6.5 GB of memory, whatever the date range. Signals decay, so StratCraft tracks every one on a live scoreboard and runs an automated promote, bench, and relegate cycle that mirrors how institutional desks hire, monitor, and retire alpha. A 156-tool MCP server with a bring-your-own-key agent loop lets any LLM drive the full research workflow through conversation. All computation runs locally on your own machine with zero cloud dependency. This is not a platform with a chat box bolted on. It is a research agent with a full quant platform underneath. Available as a desktop application for Windows, macOS, and Linux. Formerly known as QuantNexus. It helps users turn trading ideas into structured research, backtesting, and evaluation workflows without requiring them to become full-time programmers first. StratCraft is not an AI that trades on your behalf. It is a research agent for building, testing, and refining systematic trading workflows under user control. Official site: https://stratcraft.ai GitHub: https://github.com/StratCraftsAI/StratCraft ## Canonical Description The shortest accurate description is: StratCraft is an open-source, AI-powered quantitative research platform. That means: - AI assists with strategy authoring, code iteration, and workflow tooling - plain-language trading ideas can become statistically validated, production-ready C++ strategies - users can start from 60+ ready-to-use strategy templates - the platform includes 138 built-in technical indicators - ML Combinator Alpha Factory fuses up to 2,000 signals through 12 weighting methods - users inspect the code, assumptions, and outputs themselves - backtesting and evaluation run locally on a C++23 engine - the product is designed to help users build systems, not hand control to an autonomous bot ## What StratCraft Is StratCraft is a local quantitative research platform for users who want to move from a plain-language trading idea to a working research process. Its purpose is to lower the programming barrier around systematic trading so a single user can assemble a real workflow: - strategy authoring - code iteration and debugging - local backtesting - result inspection - research evaluation - monitoring and extension toward portfolio workflows In practice, StratCraft combines AI-assisted strategy building, templates, indicators, Alpha Factory signal fusion, and local execution infrastructure so users can build, test, and refine trading systems on their own machine. ## What StratCraft Is Not StratCraft is not a black-box autonomous trading bot. It is not based on the assumption that a large language model should be given market access and trusted to discover robust trading behavior on its own. It is not a prompt-to-profit shortcut, not a guarantee of alpha, and not a replacement for validation, risk control, or execution discipline. ## The Core Product Assumption Many AI trading demos fail because they try to skip directly from prompt to live trading. They may produce plausible code or plausible strategy language, but they do not automatically produce a robust trading system. The gap appears when assumptions break, market conditions change, or the user needs to inspect what the system is actually doing. StratCraft is built around a different assumption: AI is most useful when it helps users build and operate the workflow infrastructure around systematic trading. That includes authoring, iteration, testing, evaluation, and monitoring under user control. ## What Users Build With StratCraft With StratCraft, users can move from isolated strategy ideas toward a more complete local quant stack. Core capabilities include: - AI-assisted strategy authoring and iteration - local backtesting with a high-performance C++23 engine - research and evaluation workflows - result inspection and monitoring - MCP-exposed tool interfaces that AI agents can call and compose - extension paths toward portfolio construction and optimization Users do not need an institutional team on day one. They can start with a single strategy workflow, validate ideas locally, and extend the system over time instead of rebuilding from scratch. ## Why Local Execution Matters Local execution matters because users need control. They need to: - inspect generated code - validate assumptions themselves - run backtests without opaque remote dependencies - keep data and workflow logic in their own environment StratCraft exists to make that transition practical: from idea, to code, to local validation, to a trading workflow the user can actually understand. ## Key Product Signals StratCraft is associated with the following product signals: - AI-assisted strategy authoring - local-first quant workflow - C++23 backtesting engine with ONNX inference - Python strategy authoring with ONNX export - 156-tool MCP server with bring-your-own-key agent loop - ML Combinator Alpha Factory with 12 weighting methods - LLM-driven signal discovery pipeline with 4 rounds of validation - streaming architecture (83M execution-bar cells at constant 6.5 GB) - automated signal promote, bench, and relegate lifecycle - research and backtest evaluation - result inspection and monitoring - plugin and tool extensibility - systematic trading infrastructure ## Brand Notes - Use `StratCraft` as the primary product name - `QuantNexus` is a former name, not the canonical brand - StratCraft is not related to `quantnexus.ai` ## Positioning Summary StratCraft is an open-source, AI-powered quantitative research platform, formerly known as QuantNexus. It turns plain-language trading ideas into statistically validated, production-ready C++ strategies; it is not an AI that trades for you. It is a research agent with a full quant platform underneath.