tvsh

TradingView shell. Quotes, technicals, screeners, and an MCP server so Grok, Claude, Cursor, and any other tool-calling model can pull live market data.

Node 18+. No extra packages.

Install

npm install -g github:notronel/tvsh

Or run once without installing:

npx -y github:notronel/tvsh quote AAPL

From a clone:

git clone https://github.com/notronel/tvsh.git
cd tvsh
npm link

Commands

tvsh quote AAPL
tvsh chart NVDA 1h
tvsh ta TSLA
tvsh search nvidia
tvsh screener gainers
tvsh screener oversold
tvsh screener crypto
tvsh mcp
tvsh tools

screener presets: gainers losers volume oversold overbought crypto.

Connect an AI model (MCP)

This is the point of the tool. tvsh mcp speaks MCP over stdio.

Cursor — Settings → MCP → add:

{
  "mcpServers": {
    "tvsh": {
      "command": "npx",
      "args": ["-y", "github:notronel/tvsh", "mcp"]
    }
  }
}

Claude Desktop — add the same block to:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

If you used npm link / global install, this is faster:

{
  "mcpServers": {
    "tvsh": {
      "command": "tvsh",
      "args": ["mcp"]
    }
  }
}

Restart the host after saving.

Tools the model gets

Tool	What it returns
search_symbols	TradingView symbol search
get_quote	Last, volume, RSI, MACD, SMAs, oscillator rec
get_chart	OHLCV candles
run_screener	TV scanner presets

Optional: tvsh ask

If XAI_API_KEY is set, you can ask in the shell and Grok will call the same tools:

export XAI_API_KEY=...
tvsh ask "is NVDA overbought vs AVGO?"

Not required for MCP. Your editor’s model does the reasoning.

Notes

Unofficial TradingView scanner + Yahoo history. Not affiliated with TradingView. Not financial advice.