Vibe-Trading
An MCP server for agentic trading: backtesting, 462 prebuilt alphas, trade journal analysis and multi-agent research teams
High risk
We rate an entry high when the tool writes to external systems, handles money, production databases or secrets, or runs arbitrary commands. The CLI installs it only with your consent.
Why this level
- Works with financial data and can connect to real brokerage accounts
- Multi-agent teams spend external LLM tokens and can produce inaccurate recommendations
Install
Manual install
pip install vibe-trading-aiInstalls the vibe-trading CLI, the web server, and vibe-trading-mcp.
This is third-party code. Review the repository files before installing.
What it does
Vibe-Trading installs with a single pip command and starts an MCP server with dozens of quantitative research tools: market data from 28 sources, vectorized backtesting across 10 engines (China, Hong Kong, US, India, Korea, Vietnam equities, crypto, futures, forex, options), Black-Scholes option pricing and chart pattern detection. A separate toolset parses a broker's CSV export, profiles the trader's behavioral biases, and backtests rules the agent distilled from the profitable trades. For heavier research there are 30 prebuilt multi-agent teams, such as an investment committee that debates bull and bear cases.
Who it is for. For traders and analysts who want to run strategies and dissect their trades through an agent instead of manual work in Excel or one-off scripts.
Good fit when
- You need to quickly backtest a strategy on stocks, crypto or options without writing code from scratch
- You need to understand behavioral mistakes from a broker's trade export and see what would have been more profitable
- You need factor analysis or ready-made alpha factors for quantitative research
Not a fit when
- You need fully automatic live order execution without human confirmation
- You need data for markets outside the supported list of exchanges and sources
Example request
Backtest AAPL with a MACD crossover strategy (12, 26, 9) for 2024Limitations
Core tools (Hong Kong, US, Canada equities, crypto) work with no keys, but China A-shares work better with a Tushare token, and the run_swarm multi-agent teams need an LLM provider key. Connecting to real brokers (IBKR, Futu, Longbridge) needs separate setup and a local terminal session. The maintainer explicitly warns that unrelated tokens and social accounts with similar names are not affiliated with the project.
How to disable. Remove vibe-trading from your agent's mcpServers config and run pip uninstall vibe-trading-ai.
MCP
- Transport
- stdio
- Authentication
- API key
| Environment variables | |
|---|---|
| TUSHARE_TOKEN secret | Tushare token for premium China A-share data; free sources work without it |
| OPENAI_API_KEY secret | LLM provider key, needed only for the run_swarm multi-agent teams |
| LANGCHAIN_MODEL_NAME | Model name for run_swarm |
Security check
- Works with financial data and can connect to real brokerage accounts
- Multi-agent teams spend external LLM tokens and can produce inaccurate recommendations
README in short
The README presents Vibe-Trading as a personal trading agent: one pip install gives an interactive CLI, a FastAPI web server, and an MCP server. Key sections cover the Shadow Account flow (parsing a trade journal and distilling rules), backtesting across 10 engines and 28 data sources, 462 prebuilt alphas from five bundled sets (qlib158, alpha101, gtja191, academic, fundamental), 90 finance skills, and 30 multi-agent research teams. It also documents connecting external MCP servers to the built-in agent and broker connectors with separate paper and live profiles. MIT licensed, published on PyPI.
FAQ
Do I need a paid API to start?
No, Hong Kong, US, Canada and crypto data are free with no keys. Paid keys are only needed for premium sources and multi-agent teams.
Can it place real orders?
Yes through broker connectors like IBKR, but the public surface is focused on research and backtesting, and live execution needs a separate connector authorization.
Related
A set of investment research skills for Claude Code and Codex built on four investors' methodologies. It is a research tool, not investment advice
Stripe's official MCP server, plugins and skills: the agent searches the API and docs, reads and changes account data
Alpaca's official MCP server: orders for stocks, ETFs, crypto and options, position management and market data in plain language
inn-check-ru: counterparty risk traffic light
inn-check-ru
An open-source skill that checks a Russian company by tax ID across EGRUL, bailiffs, courts, bankruptcy and sanctions, and returns a dated risk verdict