Verified project record

HKUDS/Vibe-Trading

An open-source research workspace that connects natural-language prompts to market-data loaders, strategy generation, backtest engines, reports, and persistent research memory. It supports A/HK/US equities, crypto, futures, and forex backtesting, and is installable via Python, Docker, Node, source, or hosted methods.

Project overview

The project maps natural-language inputs directly to cross-market financial analysis and backtesting workflows, providing 461 cross-sectional alphas and rule-based broker journal diagnostics.

Project type: MCP · AI Agent · Data Processing
Use cases: Data Analysis · Automation
Deployment: Refer to project documentation
License: MIT

Best for

Developers and data teams needing a natural-language interface for generating, validating, and backtesting cross-market financial strategies and alphas.
Users looking to diagnose broker-journal behavior by comparing rule-based Shadow Accounts against imported trading records.

Key capabilities

Provides a self-improving trading agent for natural-language market research, strategy drafts, file/web analysis, and memory-backed workflows.
Coordinates multi-agent trading teams for investment, quant, crypto, and risk workflows, with streaming progress, persisted reports, and workers grounded with fetched market data.
Supports backtesting across A/HK/US equities, crypto, futures, and forex, with data fallback and composite backtests, PIT data validation, and run cards.
Provides diagnostics of broker-journal behavior, rule-based Shadow Account comparisons, and exportable audit reports and strategy code.
Supplies 461 cross-sectional alphas, lookahead-banned at the operator layer, with IC + IR + alive/reversed/dead categorisation in one CLI command.
Exposes tools over Model Context Protocol with stdio, SSE, and streamable HTTP transports.
Runs sessions through WebSocket, Telegram, Slack, Discord, Matrix, WhatsApp, Signal, QQ/NapCat, WeChat/WeCom, Feishu/Lark, DingTalk, Teams, email, and Mochat with CLI, REST, and Web UI controls.
Provides volatility surface, Greeks dashboard, and a payoff/scenario explorer.

Limitations and risks

MCP tool execution is serial only in v1.
Custom data loaders only support point-in-time historical bars, not real-time ticks or order-book depth.

Getting started

Install via a one-line pip install (pip install vibe-trading-ai), initialize the interactive setup (vibe-trading init), and execute a prompt (vibe-trading run -p "<prompt>"). No coding is required for this path.

Evidence and sources

GitHub project description: "Vibe-Trading: Your Personal Trading Agent"
README: Vibe-Trading is an open-source research workspace for turning finance questions into runnable analysis. It connects natural-language prompts to market-data loaders, strategy gener…
README: <b>One Command to Empower Your Agent with Comprehensive Trading Capabilities</b>
Release: v0.1.11 — India equity, fundamental factors, IM channels + roll-up
README: 🔍 Self-Improving Trading Agent

Reviewed at: 2026-07-17T10:48:24Z

Open original GitHub project