tradingview-mcp MCP — Anwendungsfälle, Installieren & Live-Demo

Warum nutzen

Hauptfunktionen

30+ indicators: RSI, MACD, Bollinger Bands, Ichimoku, etc.
Backtest 6 built-in strategies with Sharpe/Calmar/Expectancy metrics
Candlestick pattern detector (15 patterns)
Reddit + RSS sentiment (market_sentiment, financial_news)
Multi-exchange screener across 20+ filters

Live-Demo

In der Praxis

tradingview.replay ▶ bereit

0/0

Installieren

Wählen Sie Ihren Client

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

{
  "mcpServers": {
    "tradingview": {
      "command": "uvx",
      "args": [
        "tradingview-mcp"
      ],
      "_inferred": true
    }
  }
}

Öffne Claude Desktop → Settings → Developer → Edit Config. Nach dem Speichern neu starten.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "tradingview": {
      "command": "uvx",
      "args": [
        "tradingview-mcp"
      ],
      "_inferred": true
    }
  }
}

Cursor nutzt das gleiche mcpServers-Schema wie Claude Desktop. Projektkonfiguration schlägt die globale.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "tradingview": {
      "command": "uvx",
      "args": [
        "tradingview-mcp"
      ],
      "_inferred": true
    }
  }
}

Klicken Sie auf das MCP-Servers-Symbol in der Cline-Seitenleiste, dann "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "tradingview": {
      "command": "uvx",
      "args": [
        "tradingview-mcp"
      ],
      "_inferred": true
    }
  }
}

Gleiche Struktur wie Claude Desktop. Windsurf neu starten zum Übernehmen.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "tradingview",
      "command": "uvx",
      "args": [
        "tradingview-mcp"
      ]
    }
  ]
}

Continue nutzt ein Array von Serverobjekten statt einer Map.

~/.config/zed/settings.json

{
  "context_servers": {
    "tradingview": {
      "command": {
        "path": "uvx",
        "args": [
          "tradingview-mcp"
        ]
      }
    }
  }
}

In context_servers hinzufügen. Zed lädt beim Speichern neu.

claude mcp add tradingview -- uvx tradingview-mcp

Einzeiler. Prüfen mit claude mcp list. Entfernen mit claude mcp remove.

Anwendungsfälle

Praxisnahe Nutzung: tradingview-mcp

How to generate a daily market brief with signals and sentiment

👤 Retail traders, analysts ⏱ ~15 min intermediate

Wann einsetzen: Pre-market coffee, 10 minutes before the open.

Voraussetzungen

Python + pip — pip install tradingview-mcp-server

Ablauf

Snapshot the global state

Give me a market_snapshot: S&P500, NASDAQ, VIX, BTC, ETH, EURUSD, gold.✓ Kopiert

→ Table of quotes + 24h change
Scan for signals

Use scan_by_signal to find S&P 500 stocks showing RSI < 30 AND MACD bullish crossover today.✓ Kopiert

→ List of 5-30 tickers
Layer sentiment

For each of the top 5, run combined_analysis — technical + Reddit + news.✓ Kopiert

→ Per-ticker composite rating

Ergebnis: A five-ticker watchlist with a reasoned thesis per name.

Fallstricke

Reddit sentiment is a crowd signal, not a crystal ball — Never act on sentiment alone — always pair with a technical or fundamental thesis
Yahoo Finance quotes lag by 15 minutes for US stocks — For real-time intraday, confirm via your broker before trading

Kombinieren mit: alpaca

How to backtest a simple strategy before risking capital

👤 Self-directed traders ⏱ ~20 min advanced

Wann einsetzen: You want numbers, not vibes, on whether an idea works.

Ablauf

Pick a symbol and strategy

backtest_strategy: BB_bounce on SPY 2020-01-01 to 2026-04-01. Report Sharpe, max drawdown, trade count.✓ Kopiert

→ Metrics table + equity curve description
Compare strategies

compare_strategies on the same symbol and window — rank all 6.✓ Kopiert

→ Ranked table
Sanity check

Does the top strategy just ride a bull market? Show its performance only during drawdown periods.✓ Kopiert

→ Regime-specific breakdown

Ergebnis: A realistic view of whether a strategy has edge or just beta.

Fallstricke

Backtest overfitting — the best past strategy is often the worst live — Use out-of-sample windows and worry about Sharpe degradation, not peak Sharpe

How to screen crypto across Binance, KuCoin, and Bybit

👤 Crypto traders watching multiple venues ⏱ ~15 min intermediate

Wann einsetzen: You want to find setups on whichever exchange has the cleanest liquidity.

Ablauf

Screen the filter

screen_stocks with exchange in [BINANCE, KUCOIN, BYBIT]: price change 24h > 5%, volume in USD > 10M, RSI between 40 and 60.✓ Kopiert

→ Cross-exchange list
Check patterns

For each result, run get_candlestick_patterns on the 4h timeframe.✓ Kopiert

→ Patterns surfaced per ticker

Ergebnis: A short list of candidates with technical context.

Kombinationen

Mit anderen MCPs für 10-fache Wirkung

tradingview + alpaca

Screen with tradingview, then paper-trade via alpaca to validate

Find 3 SPY options setups with tradingview signals, open paper-trade positions via alpaca, and log to a Google Sheet.✓ Kopiert

tradingview + google-workspace

Log daily briefs to a Google Sheet for longitudinal tracking

Run a daily market_snapshot and append to tab 'Daily' in spreadsheet xyz.✓ Kopiert

Werkzeuge

Was dieses MCP bereitstellt

Werkzeug	Eingaben	Wann aufrufen	Kosten
market_snapshot		Morning brief	free (Yahoo)
yahoo_price	symbol: str	Real-ish-time quote (15m delayed on US equities)	free
get_technical_analysis	symbol, timeframe	RSI/MACD/BB snapshot	free
get_bollinger_band_analysis	symbol, timeframe	Mean-reversion screening	free
get_candlestick_patterns	symbol, timeframe	Pattern confirmation	free
screen_stocks	filters: {...}	Multi-factor scans	free
scan_by_signal	signal: str, universe?: str	Filter by one signal type	free
backtest_strategy	strategy: str, symbol, start, end	Historical validation	free
compare_strategies	symbol, start, end	Broad comparison	free
market_sentiment	symbol: str	Retail sentiment check	free (Reddit)
financial_news	topic?: str	Headlines feed	free (RSS)
combined_analysis	symbol	One-shot full ticker read	free

Kosten & Limits

Was der Betrieb kostet

API-Kontingent: Yahoo Finance: informal caps (roughly 2000 req/h). Reddit RSS: 60/min per IP.
Tokens pro Aufruf: Screens: 500-3000 tokens depending on result count
Kosten in €: Free — all data sources are public
Tipp: Cache indicator calls for the same bar close — rerunning inside a minute is wasteful.

Sicherheit

Rechte, Secrets, Reichweite

Minimale Scopes: None — read-only public data

Credential-Speicherung: None required

Datenabfluss: Yahoo Finance, TradingView, Reddit, RSS feeds, Binance/KuCoin/Bybit public endpoints

Niemals gewähren: Never use this for order placement; it is analysis-only

This is information tooling, not investment advice. Don't let the LLM execute trades — wire a human in the loop.

Fehlerbehebung

Häufige Fehler und Lösungen

Yahoo returns 429

Too many requests in a short window. Back off 60s and reduce request frequency.

TradingView indicator returns null

Symbol may not be listed on TradingView or timeframe is wrong. Try a known symbol like SPY first.

market_sentiment returns no data

The ticker may not be discussed on the tracked subreddits. Low-sentiment = no signal, not an error.

backtest runs forever

Narrow the date window; long histories with 1-minute bars are slow. Use daily bars first.

Alternativen

tradingview-mcp vs. andere

Alternative	Wann stattdessen	Kompromiss
yfinance MCP	You only need raw Yahoo data, no indicators	No screeners/backtests/sentiment
Alpaca MCP	You want to actually execute trades	Requires a brokerage account; different tool scope

Mehr

Ressourcen

📖 Offizielle README auf GitHub lesen

🐙 Offene Issues ansehen

🔍 Alle 400+ MCP-Server und Skills durchsuchen