Vibe Coding Discoverqanat
View on GitHubAgent-native workflow engine for building and backtesting alphas as DAGs.
Agent-first backtesting engine that turns raw data into portfolio weights via a YAML-defined DAG of SQL/Python steps over DuckDB. Ships an MCP server so agents build, run, and inspect strategies; walk-forward backtests with fees, slippage, and in/out-of-sample splits.
Use Cases
Build momentum/low-vol alpha strategies as pipeline stepsBacktest daily/weekly rebalanced portfolios with fees and slippageAgent-driven strategy authoring in plain EnglishWalk-forward replay over historical dataIn-sample vs out-of-sample performance splittingCombine multiple alphas into one priced bookForward-live scoring and scheduling of alphasInspect pipeline tables and PnL via console or MCP toolsConnect REST data sources to a DuckDB storePortfolio weight generation per strategy
Built With
- Language
- Python
- Frameworks
- FastAPI · DuckDB · pandas · NumPy · Pydantic · Uvicorn · SQLAlchemy · pytest · Hatchling · MCP
Tags
quant · backtesting · alpha-research · dag · duckdb · mcp · trading · portfolio · pipeline · yaml · fastapi · agent-native · feature-engineering · tui · cli · walk-forward