PA_Agent is an AI-driven price action analysis tool designed for traders, providing multi-source data analysis and decision support without direct trading execution.
Source: Official README and linked documentation View on GitHub →PA_Agent is being discussed because its integration of AI for price action analysis, supporting multiple data sources, and its non-intrusive approach to trading decision-making. The project differs through its two-stage AI analysis process and interactive decision tree visualization.
Source: Official README and linked documentationPA_Agent can read K-line data from MT5, TradingView, yfinance, and AkShare, offering traders a broad view of the market.
Source: Official README and linked documentationThe tool performs a two-stage analysis process, including market diagnosis and trading decision-making, use AI to assist in decision support.
Source: Official README and linked documentationPA_Agent can reuse previous analysis conclusions for new K-line data and trigger new analyses automatically upon the closing of new K-lines.
Source: Official README and linked documentationThe tool features an interactive, cyberpunk-style decision tree visualization that animates the process of gate opening to strategy path.
Source: Official README and linked documentationPA_Agent predicts the direction of the next K-line and the next market cycle position using AI.
Source: Official README and linked documentationThe tool includes a full conversation session manager with real-time reasoning and token progress bar, allowing for persistent history of dialogues.
Source: Official README and linked documentationPA_Agent provides a historical case database for analysis reference, categorized by cycle position.
Source: Official README and linked documentationThe tool logs prompts, original responses, diagnostic/decision JSON, token usage, and follow-up records for broad trading history.
Source: Official README and linked documentationPA_Agent includes a JSON validation, consistency check, semantic validation, truncation repair, and automatic retry on failure.
Source: Official README and linked documentationThe tool stores API keys locally with encryption for security.
Source: Official README and linked documentationThe architecture of PA_Agent suggests a modular design with clear separation of concerns. It uses PyQt6 for the GUI, numpy/pandas for data processing, and openai for AI API interactions. The project employs a two-stage analysis process, with data flow from data sources through preprocessing to AI analysis and decision-making.
Source: Official repository tree and dependency filesPyQt6pyqtgraphnumpypandasopenaiaksharebaostockPA_Agent is suitable for traders who want to incorporate AI-driven analysis into their trading strategy, particularly those who use MT5, TradingView, yfinance, or AkShare for data. It is useful for market analysis, decision support, and learning from historical trading cases.
Source: Official README and linked documentationAs of 2026-09-23, the project has no GitHub release. pyproject.toml reports package version 0.1.0. Because main continues to change, 0.1.0 should not be described as a published stable release.
Source: github.com/rosemarycox5334-debug/PA_Agent/releases (1) · github.com/rosemarycox5334-debug/PA_Agent/blob/main/pyproject.toml (2)status: docs-only | checked_at: 2026-09-23 | environment: Docs only. Reviewed the official README, pyproject.toml, usage and configuration documentation, AGPL license, and GitHub API. The GUI was not installed and no market-data or AI API was connected. | limitations: Dependency installation, market-data accuracy, API-key storage, model JSON, token cost, continuous tracking, backtest performance, and trading outcomes were not tested locally. | observed_output: GitHub Releases is empty. pyproject reports version 0.1.0 and Python 3.11 or newer. The README identifies Windows 10 and 11 as the primary platforms, provides a TradingView path for macOS 12 or newer, and states that the application neither connects to a broker nor places orders.
Source: github.com/rosemarycox5334-debug/PA_Agent/releases (1) · github.com/rosemarycox5334-debug/PA_Agent/blob/main/pyproject.toml (2)PA Agent is a desktop tool for one-off candlestick-chart diagnosis and manual review. Microsoft Qlib is a quantitative research platform covering data processing, model training, backtesting, risk, and portfolio workflows. PA Agent fits a quick review and record of a model's chart interpretation. Qlib fits repeatable training, backtesting, and portfolio research. Neither replaces data-quality checks or real risk controls.
Source: github.com/rosemarycox5334-debug/PA_Agent (1) · github.com/microsoft/qlib (2)PA_Agent is a project for traders looking to integrate AI into their trading strategies. Its unique features and modular architecture make it a tool for those with a technical background in AI and data analysis.
GitHub Releases is empty. pyproject reports version 0.1.0 and Python 3.11 or newer. The README identifies Windows 10 and 11 as the primary platforms, provides a TradingView path for macOS 12 or newer, and states that the application neither connects to a broker nor places orders.
The review environment was: Docs only. Reviewed the official README, pyproject.toml, usage and configuration documentation, AGPL license, and GitHub API. The GUI was not installed and no market-data or AI API was connected. PA_Agent behavior outside that scope still needs a local check.