leakguard-mcp (Model Context Protocol Server)
The io.github.doazvjettu/leakguard-mcp MCP server performs static analysis to flag lookahead bias and data leakage in machine learning and trading code prior to backtesting. It is aimed at inspecting Python code for common sources of incorrect evaluation in time-series workflows.
🛠️ Key Features
- Static analyzer focused on lookahead bias detection
- Static analyzer focused on data leakage detection
- Targets ML and trading code before backtesting
- Uses/code-related tooling aligned with libcst and pandas
🚀 Use Cases
- Pre-backtest validation of ML/trading pipelines
- Time-series model development checks
- Code-quality reviews for data leakage and evaluation bias
⚡ Developer Benefits
- Earlier identification of leakage and lookahead issues
- Helps reduce invalid backtest results caused by improper data usage
⚠️ Limitations
- Described scope is limited to static analysis performed before backtests