ZhuLinsen (open source) · Daily Stock Analysis (DSA): Open-Source AI Agent That Sends Daily Stock Watchlist Reports to Chat
Daily Stock Analysis (DSA) is an open-source (MIT) LLM agent that analyzes a stock watchlist on a schedule, combining quotes, technical indicators, news and fundamentals into scored decision reports, and pushes them to WeChat Work, Feishu, Telegram, Discord, Slack, email and other channels. It suits self-directed investors, especially those following A-shares and Hong Kong or US stocks, who want a free daily research digest they run themselves with their own LLM key.
Best for
Self-directed investors and finance-curious developers, especially those following A-shares, Hong Kong or US stocks, who want an unattended daily LLM research digest on their own watchlist delivered to a chat app, running free on GitHub Actions or their own machine with their own model key
Not ideal for
Anyone looking for investment advice or a bot that trades for them, traders who need guaranteed real-time data or execution, quant researchers who want to backtest their own trading rules, and users who need full-depth coverage of Japanese, Korean or Taiwanese markets or English-first documentation
Who it's for
Individual investors, finance hobbyists and developers who want a self-run, LLM-written daily report on their stock watchlist delivered to WeChat Work, Feishu, Telegram, Slack or email
Among the open-source LLM stock agents listed here, DSA is the one built around delivery: its core job is to turn a watchlist into a scored daily digest that lands in the chat app you already read, and the GitHub Actions setup makes that nearly free to run. The depth is uneven, though. A-shares get many data sources and market-specific context, while Japanese, Korean and Taiwanese stocks get Yahoo Finance basics, and the free data sources it starts with can be rate-limited. Its backtest scores the AI's own past calls, which is a useful sanity check, but it is not a way to test strategies. Treat the reports as a structured research starting point, configure a stable data source if you rely on scheduled runs, and check the report-backtest results for your own watchlist before trusting any score.
Who should use it
Investors who keep a watchlist of A-share, Hong Kong or US stocks and want an automated daily summary in WeChat Work, Feishu, Telegram, Slack or email; developers who want a self-hosted stock-research agent with a REST API and bots; and Chinese-speaking users who want market-specific A-share context.
Who should skip it
People who want investment advice or automated trading, quant researchers who need to backtest their own rules, users who need dependable real-time data without paying for data tokens, and anyone who needs full-depth Japanese, Korean or Taiwanese coverage or English-first documentation.
Open source
Free
Free tier limits: All features are free. The default market-data sources (AkShare, Baostock, YFinance) need no credentials but can be rate-limited and are not guaranteed to be stable.
Note: No hosted or paid plan is documented. Costs come from the LLM provider you connect and from optional paid services such as TickPlus, Tushare or Longbridge market data and SerpAPI, Tavily or other news-search keys.
API pricing
Free self-hosted FastAPI REST API (`python main.py --serve-only`); LLM usage is billed by your chosen provider
Available models
A daily watchlist digest without a server
A fork with Actions secrets runs every weekday after the close, skips non-trading days, and pushes one dashboard with a score, trend and risk alerts for each stock to a Feishu, Telegram or Slack channel.
Following up on one stock through a strategy lens
The strategy chat pulls live quotes, indicators and news and answers follow-ups from the web UI or a chat bot, using built-in strategies such as moving-average crossovers or Chan theory, or your own strategy files.
Sanity-checking AI calls before relying on them
Report backtesting compares each past call with what the price did over the following trading days and reports direction accuracy and simulated returns per stock, so you can see how the model has done on your own watchlist.
Daily Stock Analysis vs. Vibe-Trading
Vibe-Trading (MIT) is a broad research bench: natural-language backtests of your own strategies on market-specific engines, a factor library, multi-agent swarm presets, trade-journal analysis, an MCP server, and broker connectors with opt-in, capped live trading. DSA is narrower and more routine-driven: it analyzes a fixed watchlist on a schedule and pushes scored reports and alerts to chat apps, runs free on GitHub Actions, and does not trade. Choose Vibe-Trading to test ideas; choose DSA for a daily digest you do not have to run by hand.
Daily Stock Analysis vs. TradingAgents
TradingAgents (Apache-2.0) is a LangGraph framework where a fixed team of analyst, bull and bear researcher, trader and risk agents debates one ticker on one date, with grid backtesting and orders sent to a simulated exchange. DSA analyzes a whole watchlist every trading day, delivers the results to chat and email, and wraps the pipeline in a web UI, desktop app, alerts and bots. TradingAgents suits people studying or extending a multi-agent decision pipeline; DSA suits people who want the output delivered to them every day.
What is Daily Stock Analysis (DSA)?
Daily Stock Analysis is an open-source LLM agent maintained by GitHub user ZhuLinsen. It analyzes a stock watchlist on a schedule using quotes, technical indicators, news and fundamentals, writes a scored decision report for each stock and a market review, and pushes them to chat apps or email. It also has a web workspace, a desktop app, a strategy chat and a REST API.
Is Daily Stock Analysis free?
Yes. The software is MIT-licensed and there is no hosted or paid plan. You pay your own LLM provider (or run a local model through Ollama) and any optional paid data or news-search services, such as Tushare, TickPlus or SerpAPI. The recommended GitHub Actions setup needs no server of your own.
Can Daily Stock Analysis place trades?
No. It produces research reports, scores and alerts; no order placement is documented. Its Futu integration only reads account positions to use as the watchlist. The project states that its output is not investment advice.
Which markets does DSA support?
A-shares (including ETFs and registered A-share indices), Hong Kong and US stocks have the fullest coverage. Japanese, Korean and Taiwanese stocks work when you enter Yahoo Finance suffix codes such as 7203.T, 005930.KS or 2330.TW, with lighter fundamentals and no real-time guarantee.
Which LLMs does DSA work with?
Models are called through LiteLLM. The docs include presets for Gemini, Anthropic Claude, OpenAI, DeepSeek, Qwen (DashScope), Kimi, Zhipu GLM, MiniMax, Volcengine Doubao, SiliconFlow, OpenRouter, AIHubMix and Modelflare, plus any OpenAI-compatible endpoint and local models through Ollama.
Can I use DSA in English?
Yes. Setting REPORT_LANGUAGE to en produces English reports and chat replies (Korean is also available), and the web UI can be switched between Chinese and English. The README and full guide have English versions, but reports default to Chinese and several topic docs, such as market support, notifications and alerts, are only in Chinese.
Daily Stock Analysis vs TradingAgents: what is the difference?
TradingAgents is a Python research framework in which analyst, researcher, trader and risk agents debate a decision on one ticker and date, which you run from a CLI or code. DSA is an end-user routine: it analyzes a whole watchlist on a schedule and delivers the reports to chat apps, with a web UI, desktop app, alerts and chat bots around it.

HKUDS (Data Intelligence Lab@HKU)
Self-directed quant researchers, students, and technical retail investors who want a free, self-hosted agent that pulls multi-market data, writes and backtests strategies, runs multi-agent research teams, and audits their own trade history, using their own LLM key
Free
Tauric Research
Researchers, students, and technically minded investors who want an open-source, extensible multi-agent framework for studying LLM-driven market analysis and backtesting it
FreeDaily Stock Analysis (DSA) is an MIT-licensed project maintained by GitHub user ZhuLinsen. You give it a watchlist, at least one LLM key and at least one notification channel. On each run it gathers quotes, technical indicators, news and fundamentals, has the model write a scored report for every stock plus a market review, and pushes a combined decision dashboard to your chat app or inbox. The recommended setup is a fork of the repository that runs on GitHub Actions with your keys stored as secrets, by default every weekday at 18:00 Beijing time, so no server is needed. Running it locally, in Docker or as the Windows or macOS desktop app adds a web workspace with strategy chat, screening, portfolio tracking, alerts and a REST API. Coverage is deepest for Chinese A-shares, followed by Hong Kong and US stocks; Japanese, Korean and Taiwanese stocks get lighter Yahoo Finance data. Compared with research tools such as TradingAgents or Vibe-Trading, DSA is built around an unattended daily routine rather than on-demand experiments, and it does not place trades. Its default data sources are free but can be rate-limited, and the project presents its output as research, not investment advice. The repository had about 71.5K GitHub stars on November 14, 2026, and ships a release every few weeks.
v3.34.0 added optional FastAPI security middleware, an option to skip the duplicate startup analysis in scheduled mode, and offline agent trajectory diagnostics for stock-scope drift. It refreshed the default Claude, Gemini, OpenAI and DeepSeek models, prefers MiniMax M3.1 Flash for new text channels, and removed the experimental Codex App Server agent backend.
v3.33.0 added a rule-based ETF dual-momentum rotation with cost-aware backtests that does not call an LLM, a read-only data capability center, a portfolio risk and exposure dashboard, and next-action hints on the watchlist. Chat now keeps the stock under discussion on the server across refreshes and Bot/Web handoffs, and scheduled runs that time out push their completed partial results by default.
v3.32.0 brought registered A-share indices (33 entries) to the Web/API, Bot, CLI and GitHub Actions entry points, added a stock research aggregation API and data-source quality endpoints, improved Futu OpenD Hong Kong data, and moved scheduled analysis into a separate process with a hard timeout.
Are you the founder? Claim this listing →