backtest-kit

AI Trading MCP — AI skill for Claude Code

AI community

MCP-driven crypto trading rig: Claude (or any MCP client) trades a live portfolio through three guarded tools: get_status, open_position, close_position.

How to install AI Trading MCP

This entry records only its repository, not the path inside it, so there is no exact command to give. Open backtest-kit/ai-trading-mcp and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What AI Trading MCP does

MCP-driven crypto trading rig: Claude (or any MCP client) trades a live portfolio through three guarded tools: get_status, open_position, close_position. Capable to read news and charts. Paper and live.

Alternatives in AI

BrowserSkill — Let AI agents use your real, logged-in browser without interrupting your work 9.1k ★
Sentrux — Real-time architectural sensor that helps AI agents close the feedback loop, enabling recursive self-improveme 2.9k ★
Opennews MCP — News Aggregation · AI Ratings · Trading Signals · Real-time Updates 2.1k ★

README

👾 ai-trading-mcp

An AI news-trading rig built on [backtest-kit](https://github.com/tripolskypetr/backtest-kit): a GramJS scraper pipes a live Telegram channel — text **and chart screenshots** — straight into the MCP `get_status` feed, Claude decides *when and which way* through three guarded tools, and the trading engine owns every level, limit and validation — from paper fills to a crash-safe Binance spot broker adapter.

![screenshot](https://raw.githubusercontent.com/tripolskypetr/backtest-kit/HEAD/assets/screenshots/screenshot16.png)

[![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/tripolskypetr/backtest-kit) [![npm](https://img.shields.io/npm/v/@backtest-kit/mcp.svg?style=flat-square)](https://npmjs.org/package/@backtest-kit/mcp) [![TypeScript](https://img.shields.io/badge/TypeScript-5.0+-blue)]()

The LLM is the **only signal source** here — the bundled strategy registers no entry logic at all — and the engine is the **only executor**: the agent's whole vocabulary is `get_status`, `open_position(symbol, position, note)` and `close_position(symbol, note)`. Take-profit, stop-loss and entry cost are computed engine-side and cannot be overridden by the model. What this project adds on top of stock backtest-kit is the **sensory organ** (a Telegram feed rendered into MCP text + image blocks), the **audit trail** (every byte the model saw, dumped as markdown), and a **battle-hardened spot broker** for going live.

📚 **[API Reference](https://backtest-kit.github.io/documents/example_02_first_backtest.html)** | 🌟 **[Quick Start](https://github.com/tripolskypetr/backtest-kit/tree/master/example)** | 📰 **[Article: AI News Trading Signals](https://backtest-kit.github.io/documents/article_07_ai_news_trading_signals.html)**