AI Stock：用 AI 智能体做股票研究、策略筛选与模拟交易

AI Stock

Empower every stock researcher with AI—so one person can research with the capabilities of a team.

Your AI workspace for researching stocks, comparing ideas, and testing strategies with simulated money.

Research assistant · Expert roundtable · Stock research · Strategy screening · Trade simulation

English · 简体中文 · 繁體中文 · 日本語 · 한국어

AI Stock brings market data, news, calculations, and AI analysis into one website. Ask about a stock, read the evidence and risks, compare different investing perspectives, and test a strategy through paper trading. Research supports mainland China, Hong Kong, the US, Taiwan, Japan, Korea, the UK, Canada, Australia, India, Germany and France. Paper simulation supports the first six markets. Data availability and bounded screening coverage depend on the provider and configuration; see the market guide.

Trade simulation demo. The workspace shows seven-module top navigation, strategy controls, return comparisons and expandable paper accounts. This screenshot was captured locally with synthetic DEMO data; the rising curves do not represent actual returns. Explore the public demo.

Try it online (recommended)

The easiest way to try AI Stock is the hosted website. Register with an invitation code for a daily platform allowance, or register without one and configure your own LLM API in My account. No local deployment needed.

Open AI Stock · Request an invitation code

Click Request an invitation code to open your configured default email app with the recipient and subject filled in, then send your request. You will receive a code by email; use it to register on the website, then log in. Already have an account? Log in directly.

Use your own email account: Gmail, Outlook, QQ Mail, 163 Mail, and other providers can all send the request. If no email app opens, right-click or long-press the request link and copy the email address, then compose a message in your usual mailbox with the subject “AI Stock invitation code request”. If you copied the full link, use only the address between mailto: and ? as the recipient.

Start here: What you can do · How it works · Quick start · Your first report

Choose your language: Use the language selector at the top of the website. Interface labels follow your selection; saved reports, news and user-written content keep their original language. Report output language is configured separately.

What you can do

Example question: “Research AAPL's recent price trend and important news. Separate bullish and bearish evidence, include the dates of the data, and tell me what information is missing.”

Built-in full-market screening rules mainly target A-shares. The “experts” are AI roles based on investment frameworks, not the actual people. Trading experiments use simulated funds and do not place live broker orders.

Select the Crypto market in the existing modules; Binance Spot uses the same saved strategy and simulation workflows, with completed UTC daily bars. It does not place exchange orders. See the crypto spot guide.

How it works

1. You set the question or stock scope. Start with a chat, a stock code, or screening conditions.
2. The system gathers available evidence. Data sources supply prices, fundamentals, and news as supported; tools query data and calculate indicators.
3. AI applies the selected method. It uses the supplied evidence and strategy instructions to form an analysis. Optional experts examine the topic independently.
4. You review the result. Revisit saved reports, inspect risks and missing information, or run a separate trading simulation.

An Agent is the AI assistant doing the work. A Skill is its analysis playbook. Tools perform queries and calculations; MCP connects external tools. You can start with enabled defaults and learn these settings later.

Research and simulation have different inputs: a research report can use news and fundamentals when available, while trading decisions use the configured strategy, dated market bars, and simulated account state. See simulation inputs and limits.

Quick start

For the quickest start, use AI Stock online and request an invitation code. The installation steps below are only for running your own instance.

1. Prepare your computer

Install Python 3.10+, Node.js 20.19–26.x, npm 10+, and Git. Prepare a model provider's API key and access to a model that supports tool calling. Hosted model calls may use paid provider credits; see model configuration for supported setup options, including local models.

These commands use a macOS/Linux shell. Windows and Docker users can follow the deployment guide.

2. Download and install

git clone https://github.com/EthanAlgoX/AIStock.git AI-Stock
cd AI-Stock

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt

# Preserve any existing configuration.
if [ ! -f .env ]; then cp .env.example .env; fi

Already have the repository? Start from cd AI-Stock in its parent directory and keep your existing .env.

3. Build the website and start the server

cd apps/dsa-web
npm ci
npm run build
cd ../..

python main.py --serve-only --host 127.0.0.1 --port 8000

Open http://127.0.0.1:8000. Keep this terminal running while you use the local instance. Port 8000 is an example; if it is occupied, change --port and use the same port in your browser. API documentation is at http://127.0.0.1:8000/docs.

The website opening is only the first step. Configure and test your model before asking it to analyze stocks.

In another terminal, check curl --fail http://127.0.0.1:8000/api/health. A successful response verifies the local API; model and data-source connections need their own checks.

Your first report

Using the hosted website? Invited accounts can use their available platform allowance. Accounts registered without an invitation must first save a personal LLM API in My account and verify it with a short assistant question. Then follow steps 3–4. Instance-wide Settings and data-source administration are for the instance owner.

1. Connect AI. In Settings → Models & runtime, choose your provider, enter its API key and model details, save, and test the connection. You can also configure the provider in .env; see the model guide.
2. Check available capabilities. In the Capability Center, check enabled Skills, tools, and data sources. Configure a news source if your question needs recent news. A configured source is not necessarily reachable: use its availability check.
3. Run one stock. Open Stock research, click New stock research, choose a market, and enter a code: 600519 for an A-share, hk00700 for a Hong Kong stock, or AAPL for a US stock, then select the matching stock from the results. Review the selected research method, enter your research goal, and click Run stock research. The system supplies the built-in research configuration; keep its required Skills and tools enabled.
4. Read the saved report. Check the conclusion, supporting evidence, data dates, and risks. Missing data or partial results matter. Reopen the result from research history or Tasks & Runs. Use Research assistant for follow-up questions, or Expert roundtable to compare views.

For a first goal, try: “Explain this stock's recent trend and key risks. Show the evidence and flag anything you cannot verify.” Stock codes here are input examples, not recommendations.

The header lets you switch between English, Simplified Chinese, Traditional Chinese, Japanese, and Korean. Changing the interface language does not translate existing reports.

If something does not work

Go further when you are ready

• Compare viewpoints: choose experts and pipeline, debate, or voting mode in Expert roundtable. Collaboration guide (Chinese)
• Test a strategy: choose a Skill in Trade simulation, preview and confirm the stock scope, save, and run once or start daily simulation. Review decisions, costs, and simulated fills. Trading guide
• Try JEV decisions (optional): configure its separate TypeSafe API key in Settings → Models & runtime, then select JEV · Decisions only for a new trading strategy. JEV returns buy/sell/hold, probabilities, and confidence without a research report; chat and stock-scope selection still need your LLM. Start with the official access information, TypeSafe console, and JEV setup guide.
• Run tasks on a schedule: keep the server running, locally or on a deployed machine. Your browser can close; a local server stops when its computer is off. Deployment guide
• Track holdings or a watchlist: record holdings or followed stocks in Portfolio, save a tracking plan, and enable its schedule. Saving a plan and running research are separate actions. Workflow guide (bilingual)

Documentation and development

For frontend development, keep the backend running and run npm run dev in apps/dsa-web. Vite defaults to port 5173 and proxies /api to backend port 8000; use DSA_WEB_API_PROXY_TARGET if the backend address differs.

# From the repository root
python -m pip install -r .github/requirements-ci.txt
./scripts/ci_gate.sh

cd apps/dsa-web
npm test -- --maxWorkers=2
npm run lint
npm run build

License and scope

MIT License. AI Stock is for investment research and controlled historical or paper-trading experiments. Reports and simulated performance do not guarantee future returns. Historical AI replay can be affected by knowledge in model training; it is not proof of a profitable strategy.

Previously named LLM TradeBot and InvestCrew; the current product name is AI Stock and the repository is EthanAlgoX/AIStock.