Vibe-Trading
Multi-agent finance workspace with an alpha library and an MCP server
🧩 Build it Free 02 in Bots & automation 🆕 NEW TODAY
33,123 0 copies · 7 days MITupdated 9 Sep '26
Overview
Vibe-Trading is a research workspace that turns natural-language prompts into market research, strategy backtests, and reports, with optional broker trading you authorize yourself.
- Fortraders who want an agent to research, backtest, and report on strategies across equities, crypto, futures, and forex.
- Needsan LLM API key from a supported provider (or local Ollama with no key), Python 3.11+ for a local install, or Docker.
- Runsa Python CLI/TUI, a FastAPI web server, and an MCP server (vibe-trading-mcp) for plugging into an existing agent like Claude Desktop or Cursor.
- Limitits broker-trading capability is experimental and not verified by the maintainers against a real broker account.
Vibe it
Paste this into your AI agent. Write the first line, or take a suggestion; the copy carries your words.
Source: https://github.com/HKUDS/Vibe-Trading — clone it and read its README and examples before writing anything; its documented setup beats what you remember. Stack: a Python agent (FastAPI backend, React frontend) with market-data loaders, backtest engines, persistent research memory, and an MCP server for plugging into an existing agent. Needs: an LLM API key from a supported provider (OpenAI, Anthropic, DeepSeek, Gemini, and others), or local Ollama with no key; Python 3.11+ or Docker. Start: `pip install vibe-trading-ai`, then `vibe-trading init` to set up my .env, then run `vibe-trading run -p "Backtest a BTC-USDT 20/50 moving-average strategy for 2024 and summarize return and drawdown"`. Limit: its broker-trading capability is experimental and not verified by the maintainers against a real broker account. Done means: you show me the backtest's return and drawdown summary for the example strategy, and the report file it produced. Never place a live order. Paper or testnet until I say live in this chat. Before you start, ask me: which LLM provider I'm using (the key goes in the env file, never in this chat), and which market and strategy idea I want backtested first.
