Purpose
TradingAgents-CN is a Chinese-language financial trading research framework based on multi-agent LLMs. The community edition is oriented toward the Chinese A-share market and is positioned as a multi-agent stock research assistance system. The project aims to help users systematically learn the application of multi-agent frameworks and AI large models in stock research. It should be specially noted that this project is a research assistance tool; all AI-generated content is for learning reference only and does not constitute investment advice.
Main Capabilities
- Single-stock research: Uses multi-agent debate-style analysis, supports 5 levels of depth, and presents multi-scenario viewpoints.
- General research: Research any market topic by asking questions in natural language.
- Stock screening: Combine natural language with conditional screening to discover targets from 6,000+ stocks.
- Stock selection assistant: Match candidate stock pools according to research requirement descriptions.
- Intelligent assistant: Web search + tool invocation + factual memory.
- AI memory: Project-level memory and fact accumulation.
- Skill Center: Create and manage custom data acquisition skills.
- Learning Center: Provides free systematic courses (5 lessons in the discovery phase) + 8 learning guides.
The system completes the research process through collaboration among multiple AI agents playing different roles (market-wide/sector/market/fundamental/news analysts, optimistic/prudent researchers, etc.).
Usage
Environment Dependencies
You need to install and start MongoDB 7.0+ (8.0 recommended) and Redis 6+ (7 recommended) yourself; the project does not include a database. You can also use Docker to run only these two databases (Docker must already be installed locally). MongoDB must have authentication enabled (authSource=admin), and you must create an administrator account yourself; write the username/password, host, and port into the .env file (local defaults are localhost:27017 / localhost:6379), and keep them consistent with the database account password and port.
Initialization and Configuration
On first startup, the backend automatically completes initialization without manually executing any import scripts: it imports system-level configurations (model catalog, prompt templates, Agent configuration, workflows, etc.); when the user table is empty, it creates the administrator account admin using ADMIN_DEFAULT_PASSWORD in .env. After logging in, please first configure the large model API Key in system settings (OpenAI / DeepSeek / Tongyi Qianwen, etc.), and configure market data Token as needed.
Data Sources
Market and financial data come from public interfaces such as Tushare / AKShare / BaoStock, and you need to register yourself to obtain a token.
Documentation and Courses
User manual: docs/02-user-guide/user-manual-v3.0.md; learning guide: docs/02-user-guide/ (learning-01 ~ 08); free courses: docs/07-courses/v3.0/phase-01-discover/.
Version and Licensing
The Community Edition uses a hybrid license: Apache-2.0 open source components include tradingagents/ (trading agent library), cli/, docs/, tests/, etc., which can be freely used, modified, and distributed; proprietary components (Source Available) include app/ (backend application), frontend/ (frontend application), core/ (core functionality layer). The source code is visible, and personal use, evaluation, and educational use are permitted, while redistribution and commercial use are prohibited (for commercial licensing, please contact the author). The Pro version is aimed at users with more advanced research workflow needs; for how to obtain it, see the project homepage or contact the author. By submitting a contribution, you agree to sign the Contributor License Agreement (CLA).
Applicable Scenarios
Suitable for developers and researchers who want to learn multi-agent frameworks and study the application of AI large models in A-share stock research, as well as individual users who need stock research assistance analysis, stock screening, and thematic research.
Limitations
This system only supports A-share market research; the output is AI-generated research reference and does not constitute investment advice; it does not provide trading instructions such as target prices, stop-loss and take-profit levels, or buy and sell timing. Users need to install and maintain MongoDB and Redis themselves, and obtain large model API Key and market data Token themselves.
Community Contributions
The project welcomes open source talent to participate. The way to join is to directly submit a Pull Request. Resumes and interviews are not valued, and PRs are the only method of assessment. Evaluation criteria include: solving real problems, explaining the reasons for changes, and responding to review comments. Partners who continuously contribute high-quality PRs will participate in version planning and key module development. Directions you can participate in include data source adaptation, frontend development (Vue 3 + TypeScript), Agent orchestration and prompt engineering, and documentation and course building. Contributors can receive public attribution and traffic, practical experience in multi-agent systems engineering, and priority in project commercialization. The project's current revenue is not yet enough to cover basic costs, so participation is temporarily voluntary; the distribution principles after revenue covers costs have been written in advance. For initial communication, you can email hsliup@163.com; for specific technical questions, it is recommended to directly submit an Issue / PR for public discussion. For the submission process, distribution framework, and Contributor License Agreement (CLA), see CONTRIBUTING.md. For complete acknowledgments, see ACKNOWLEDGMENTS.md.
