Nova-TradingAgent

What is Nova-TradingAgent

Nova-TradingAgent is a self-hosted A-share research desk that coordinates multiple AI agents around stock-market analysis. The project description highlights a 15-agent debate process, which suggests the system is designed to compare viewpoints before forming a conclusion.

It focuses on the Chinese A-share market and is distributed under AGPL-3.0. The app also points to a translation layer that converts market data into conclusions the model can reason over, instead of forcing the model to parse raw dumps directly.

In practical terms, this is closer to an AI research workstation than a brokerage platform. It appears to support analysis workflows, not order execution.

Key features

• 15-agent debate workflow: the site describes fifteen specialists working together, which implies a structured multi-agent research process rather than a single-chatbot setup.
• Self-hosted deployment: the app is meant to be installed and run by the user, so teams can keep the system in their own environment.
• A-share focus: the workflow is centered on A-share research, making it relevant for users who follow mainland China equities.
• Optional Tushare Level-2 input: the site mentions Tushare Level-2 as opt-in, so advanced market data is available only if you connect it.
• K-line workstation with AI insight: chart-based analysis is part of the experience, with AI guidance layered onto candlestick data.
• Translation layer for models: the system turns market outputs into model-friendly conclusions, which can help reduce noise from raw market feeds.
• Open-source license: AGPL-3.0 is explicitly listed, which is important for teams evaluating reuse and deployment terms.

How to use it

1. Review the project links and docs first. The site exposes GitHub, Chinese-language materials, installation instructions, and a user guide, so setup should start with the official documentation.
2. Deploy it in your own environment. Because Nova-TradingAgent is self-hosted, you should expect to manage the runtime, configuration, and any required data access yourself.
3. Decide whether to enable Tushare Level-2. The app notes that this source is optional, so you can start with the built-in workflow and add higher-detail data later if your use case needs it.
4. Use the K-line workstation for chart review. The interface appears built for candlestick-based analysis, making it useful when technical context matters.
5. Run the multi-agent research flow. The 15-agent setup is the core of the product, so use it when you want multiple model perspectives before drawing a thesis.
6. Read the model conclusions, not just the data. The translation layer is intended to present distilled findings that models can interpret more reliably than raw market tables.

Who it is for

Nova-TradingAgent is a fit for users who already work with A-share research and want a more systematic, AI-assisted process.

• Quantitative or discretionary traders who need a repeatable research desk for idea generation.
• Independent analysts who want multi-agent discussion instead of single-model summaries.
• Self-hosting teams that prefer local control over infrastructure and workflows.
• Developers and power users who are comfortable reading installation docs and working with API-driven market data.
• Researchers focused on Chinese equities who want chart analysis plus model-assisted interpretation in one place.

It is less suited to beginners looking for a simple stock-picking app, and it does not present itself as a consumer brokerage product.

What to know before you use it

Nova-TradingAgent has a few important limits that are worth checking before you rely on it.

• It is not a trading venue or broker. The available material describes research and analysis, not live order placement.
• Tushare Level-2 is optional. If you do not connect it, some deeper market-data workflows may not be available.
• Accuracy depends on your data setup. Self-hosted research tools are only as good as the feeds and configuration behind them, so verify every output against your own process.
• It is A-share specific. The product is framed around mainland China equities, so it may not suit users who mainly study U.S. or global markets.
• The translation layer is a helper, not a source of truth. It helps models process conclusions, but it does not replace checking the original market data when precision matters.
• Official docs matter for setup. Because installation and usage details can change, use the project’s GitHub, Chinese docs, install guide, and user guide for the latest instructions rather than relying on third-party summaries.

For teams that want a self-hosted stock research workspace with multiple AI agents, Nova-TradingAgent offers a focused A-share workflow. If your research depends on controlled deployment, chart review, and optional Level-2 data, it is worth evaluating as an open-source AI market analysis desk.