PanWatch is a self-hosted AI stock-monitoring assistant covering A-shares, Hong Kong, and US markets. Built on the TradingAgents multi-agent research framework and starred around 2,200 times on GitHub, it watches your portfolio around the clock, analyzes positions with LLM agents, and pushes real-time alerts and automated reports.
Instead of staring at charts, PanWatch’s agents do the watching for you. It tracks holdings and watchlists, generates portfolio insight reports, and sends notifications when something moves. The system is built with FastAPI and LangGraph, supports MCP tool integration, and runs as an installable PWA, so your market data and LLM keys stay on your own server.
Market data comes through open data libraries such as AKShare for Chinese markets, and the analysis layer builds on TradingAgents, the multi-agent framework in which specialized analyst agents debate market conditions before producing a view. Reports can be scheduled โ for example, a pre-market briefing or a post-close portfolio summary โ so the system works more like a research assistant on a timetable than a dashboard you must remember to check.
PanWatch is designed to run on a home server or NAS with Docker, and its PWA front end makes alerts and reports readable on a phone without installing an app. Users can define watchlists separate from actual holdings, making it equally useful for tracking stocks they are considering buying, and report templates can focus on technical signals, fundamentals, or news-driven events.
PanWatch is free and open source under the MIT license. You supply your own market-data access and LLM API keys (for example DeepSeek or OpenAI).
Retail investors and quant-curious developers who want an AI assistant to monitor their portfolio and summarize what’s happening across markets.