Vibe-Trading: what it is, what problem it solves & why it's gaining traction

What it solves

Vibe-Trading provides a comprehensive framework for creating personal trading agents. It bridges the gap between high-level investment hypotheses and actual execution by automating market data retrieval, signal generation, backtesting, and trade attribution, reducing the manual effort required to research and deploy trading strategies.

How it works

The system operates as an AI agent powered by LLMs (supporting providers like DeepSeek, Gemini, and Kimi) that can interact with a wide array of tools. It uses a loader registry to fetch normalized market data from various sources (including Yahoo Finance, Finnhub, and local CSV/Parquet files) and can generate "SignalEngines" to test hypotheses. The "Research Autopilot" loop automates the process of moving from a hypothesis to a signal engine and then to a backtest, feeding metrics back into the hypothesis for refinement. It also includes a "Shadow Account" feature to extract and codify trading rules from past performance.

Who it’s for

It is designed for traders, quantitative researchers, and developers who want to leverage LLMs to automate the research, validation, and execution of trading strategies across A-share, US, and HK markets.

Highlights

• Extensive Data Integration: Supports 18+ market-data sources and 18 read-only data tools for fund flow, financials, and options chains.
• Research Autopilot: An end-to-end loop that automates the Hypothesis → Signal-Engine → Backtest workflow.
• Multi-Broker Connectivity: Includes 10 broker connectors for live trading capabilities.
• Advanced Attribution: Performs layered attribution after backtests, including beta regression, market-regime analysis, and Monte Carlo permutation tests.
• Swarm Intelligence: Supports "investment-committee" runs where multiple swarm workers collaborate on research tasks.