Alpaca MCP — Anwendungsfälle, Installieren & Live-Demo

Warum nutzen

Hauptfunktionen

Works against paper OR live accounts via the same tools
Stocks, ETFs, crypto, and options supported
Historical bars + real-time quotes from Alpaca Market Data
Full order lifecycle: place, replace, cancel, check status
Positions, account equity, and P/L available as first-class tools

Live-Demo

In der Praxis

alpaca.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": {
    "alpaca": {
      "command": "uvx",
      "args": [
        "alpaca-mcp-server"
      ]
    }
  }
}

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

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "alpaca": {
      "command": "uvx",
      "args": [
        "alpaca-mcp-server"
      ]
    }
  }
}

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

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "alpaca": {
      "command": "uvx",
      "args": [
        "alpaca-mcp-server"
      ]
    }
  }
}

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

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "alpaca": {
      "command": "uvx",
      "args": [
        "alpaca-mcp-server"
      ]
    }
  }
}

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

~/.continue/config.json

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

Continue nutzt ein Array von Serverobjekten statt einer Map.

~/.config/zed/settings.json

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

In context_servers hinzufügen. Zed lädt beim Speichern neu.

claude mcp add alpaca -- uvx alpaca-mcp-server

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

Anwendungsfälle

Praxisnahe Nutzung: Alpaca

Generate a daily pre-market brief for your watchlist

👤 Active retail traders ⏱ ~10 min beginner

Wann einsetzen: You want a 30-second morning check on your 20 tickers before the open — overnight moves, earnings, obvious news.

Voraussetzungen

Alpaca account (free paper) — alpaca.markets → sign up → API keys on the dashboard
Watchlist of symbols — Plain list like [AAPL, MSFT, NVDA, ...]

Ablauf

Pull overnight bar data

For my watchlist [symbols], get yesterday's close and the latest pre-market quote. Compute % gap.✓ Kopiert

→ Table: symbol, prev_close, pre_open, gap_pct
Flag outliers

Which tickers are gapping more than 2% in either direction? Rank by gap magnitude.✓ Kopiert

→ Short 'needs attention' list
Add context

For each outlier, pull the latest news headlines from Alpaca news endpoint.✓ Kopiert

→ Per-ticker bullet with likely cause

Ergebnis: A 1-screen daily pre-market brief, same shape every day, no manual clicking.

Fallstricke

Free market data is IEX-only and lags; quotes may not match your broker — Fine for overnight/EOD analysis; if you need SIP feed, upgrade to Alpaca's paid data plan

Kombinieren mit: notion

Paper-trade a simple rules-based strategy

👤 Algo-trading beginners ⏱ ~30 min intermediate

Wann einsetzen: You have a rule like 'buy SPY when 5-day return < -3%, sell after +2% or 5 trading days' and want to run it live on paper money.

Voraussetzungen

Paper account keys configured — Use APCA_API_BASE_URL=https://paper-api.alpaca.markets
A written rule — Entry, exit, position size, max loss — all explicit before coding anything

Ablauf

Check account and positions

Show my paper account equity, buying power, and current open positions.✓ Kopiert

→ Clean account snapshot
Evaluate entry condition today

Get SPY's last 5 daily bars. Compute 5-day return. If it's < -3% and I don't already hold SPY, place a market buy for $5000 notional. Otherwise do nothing.✓ Kopiert

→ Either an order id or an explicit 'no action'
Log today's decision

Append a row to /trades/journal.csv: date, condition_met, action, order_id, fill_price.✓ Kopiert

→ Journal row written

Ergebnis: An auditable daily run of a simple strategy against paper money, with a journal you can review.

Fallstricke

Accidentally running against live keys during testing — Keep a very obvious check at the start: confirm the account number matches your paper account before placing any order
Claude forgets yesterday's position and double-buys — Always fetch current positions before evaluating entry; never rely on memory

Kombinieren mit: filesystem

Rebalance a portfolio to target allocations

👤 DIY investors with a target allocation ⏱ ~20 min intermediate

Wann einsetzen: Quarterly: bring 60/30/10 (US/INTL/bonds) back to target without manually computing share counts.

Voraussetzungen

Target allocation defined — e.g. VTI 60%, VXUS 30%, BND 10%

Ablauf

Measure drift

Get my positions and current prices. Compute current weight per symbol vs target [paste targets]. Show drift in $ and %.✓ Kopiert

→ Per-symbol drift table
Propose trades

Propose the minimum set of buys/sells to bring weights within 1% of target. Show expected commissions (likely $0) and estimated slippage.✓ Kopiert

→ Order plan
Execute with confirmation

Before placing: summarize the plan in 3 lines and wait for my 'yes'. On yes, submit as market-on-open orders.✓ Kopiert

→ Orders placed with ids; equity matches target within 1%

Ergebnis: A clean quarterly rebalance with a written record of why each trade happened.

Fallstricke

Taxable account: every rebalance triggers gains/losses — Prefer new-contribution rebalancing or tax-lot-aware exits; run only in IRA if unsure

Kombinieren mit: filesystem

Scan an options chain for short-dated put premium

👤 Option income traders (cash-secured puts / covered calls) ⏱ ~25 min advanced

Wann einsetzen: You want to find the best cash-secured put to sell on tickers you'd be happy to own, weekly.

Voraussetzungen

Options-enabled Alpaca account — Enable in account settings; paper works for learning

Ablauf

Fetch chain + greeks

Get the options chain for [ticker] expiring next Friday, puts only, strikes within 5% of spot. Include bid, ask, mid, delta, IV.✓ Kopiert

→ Chain table
Rank by annualized yield

Compute annualized return-on-cash for each strike assuming assignment. Filter delta between -0.15 and -0.30. Rank top 5.✓ Kopiert

→ Ranked short list
Place sell-to-open with a limit at mid

Place a sell-to-open limit order for 1 contract of the #1 strike at the mid. GTC, day-only.✓ Kopiert

→ Order id returned

Ergebnis: A repeatable weekly income routine with a conservative risk screen.

Fallstricke

Selling puts on companies you wouldn't want to own — Hard rule: only trade this strategy on 10 tickers you've pre-approved as long-term holds
Illiquid strikes with huge bid-ask — fills never happen at mid — Filter for open_interest > 500 and spread < 5% of mid before ranking

Kombinieren mit: notion

Kombinationen

Mit anderen MCPs für 10-fache Wirkung

alpaca + notion

Daily pre-market brief delivered to a Notion page

Run my pre-market scan for watchlist and create a Notion page in 'Trading / Daily' with the results.✓ Kopiert

alpaca + filesystem

Trade journal: append every order and reason to a local CSV

After any place_order, append {ts, symbol, side, qty, reason} to /trading/journal.csv.✓ Kopiert

alpaca + postgres

Store daily equity curve for backtesting strategies

Every evening, INSERT today's account equity into the equity_curve table.✓ Kopiert

Werkzeuge

Was dieses MCP bereitstellt

Werkzeug	Eingaben	Wann aufrufen	Kosten
get_account		First call in any session — confirm you're on paper vs live	free
get_positions		Check current holdings before any entry logic	free
get_bars	symbol, timeframe, start, end, limit?	Historical OHLCV for analysis and signals	free (IEX) or paid (SIP)
get_latest_quote	symbol	Real-time price before placing an order	free
place_order	symbol, qty\|notional, side, type, time_in_force, limit_price?	Submit a new order (always paper-first)	free (commission-free)
cancel_order	order_id	Cancel an open order before it fills	free
get_options_chain	underlying, expiration, side?	Scan strikes for an options strategy	free on Alpaca

Kosten & Limits

Was der Betrieb kostet

API-Kontingent: 200 req/min on data API (free tier); trading API is effectively unthrottled at retail scale
Tokens pro Aufruf: Account/positions: 200–500 tokens. Bars: scales with limit.
Kosten in €: Alpaca is commission-free. Paid market data plan ~$9/mo for SIP feed; free IEX feed is fine for most retail use.
Tipp: Paper-trade every new strategy for at least 4 weeks before touching live money. It's free.

Sicherheit

Rechte, Secrets, Reichweite

Minimale Scopes: Account API key + secret — Alpaca doesn't have sub-scopes

Credential-Speicherung: APCA_API_KEY_ID, APCA_API_SECRET_KEY, and APCA_API_BASE_URL in env vars

Datenabfluss: All calls to api.alpaca.markets or paper-api.alpaca.markets

Niemals gewähren: Sharing live keys with any third-party service, including code hosts

LIVE keys can buy and sell real securities. Default to PAPER URLs (paper-api.alpaca.markets) until strategies are proven.
No 2FA on API keys — treat them like a password manager entry and rotate after any suspected leak.

Fehlerbehebung

Häufige Fehler und Lösungen

403 forbidden.insufficient-permissions

Your keys are for the wrong environment (live vs paper). Match APCA_API_BASE_URL to the key pair.

Prüfen: curl -H 'APCA-API-KEY-ID: ...' -H 'APCA-API-SECRET-KEY: ...' $APCA_API_BASE_URL/v2/account

422 buying_power insufficient

Your paper account reset, or you sized the trade beyond cash. Call get_account and recompute position size.

Bars return empty for recent timestamp

IEX feed has ~15-min delay on free tier, and doesn't serve every symbol. Use SIP or wait.

Order rejected: market closed

Default TIF is day; market-closed orders need time_in_force='opg' or 'cls' or switch to extended hours with extended_hours=true on limit orders.

Alternativen

Alpaca vs. andere

Alternative	Wann stattdessen	Kompromiss
Interactive Brokers MCP	You need international markets or futures	More complex auth (gateway app required); learning curve
Polygon.io MCP	You only need market data, no trading	No order execution; paid for full coverage

Mehr

Ressourcen

📖 Offizielle README auf GitHub lesen

🐙 Offene Issues ansehen

🔍 Alle 400+ MCP-Server und Skills durchsuchen