What is Vibe-Trading?
Vibe-Trading is a finance research agent that combines language-model workflows with market analysis and backtesting. Its research features include factor evaluation, correlation analysis, portfolio snapshots and retained backtest artifacts. Swarm workflows coordinate agent work while keeping observations and results available for inspection.
The project supports self-hosting through Docker, with a Web interface served by its API server and a CLI entry point. Its container configuration persists sessions, research runs, uploads and agent state. MCP exposes research tools to external clients, while saved paper accounts simulate orders or strategies against completed daily bars without sending those orders to a broker.
Research outputs still require review of market assumptions, data availability and execution settings. The published release documents issues where retried provider errors can halt live model decisions and budget exhaustion can discard a completed research answer; fixes are identified as unreleased.
Key capabilities
- Backtesting Runs strategy backtests and retains run cards, metrics and other artifacts for inspection.
- Factor research Supports factor evaluation, neutralization and principal component analysis, with checks for insufficient observations.
- Portfolio analysis Reuses immutable portfolio snapshots and exports sanitized snapshot data from the Web interface.
- Paper trading Maintains saved simulation accounts that advance through completed daily bars without reaching a broker.
- Agent workflows Coordinates Swarm runs, preserves tool observations and tracks language-model usage.
- Programmatic access Exposes research functionality through a REST API and MCP tools.
Where Vibe-Trading fits
Vibe-Trading fits finance research workflows that need to move between an agent conversation, strategy backtests and inspectable results. Researchers can examine factors and correlations, reuse portfolio snapshots and simulate daily trading before considering a separate execution workflow.
Before deployment, operators should verify model-provider access, market-data availability, engine assumptions and credential handling. They should also review network exposure and persistent storage: the supplied Compose configuration publishes locally and preserves agent state in volumes. Published-release behavior should be distinguished from fixes and features marked unreleased.
Install
pip install -U vibe-trading-aiReviews of Vibe-Trading
Be the first to share your experience with Vibe-Trading.
Write a review
Loading review form…