> ## Documentation Index > Fetch the complete documentation index at: https://mathematicalcompany.mintlify.site/llms.txt > Use this file to discover all available pages before exploring further. # MCP Portfolio Assistant > Use the Horizon MCP server to build an AI-powered portfolio management workflow in Claude Code. This example shows a complete workflow using the Horizon MCP server through Claude Code: discovering markets, analyzing opportunities with Kelly criterion, placing trades, setting protective orders, and monitoring positions - all through natural language. ## Setup ```bash theme={null} # Install pip install horizon-sdk[mcp] # or: uv pip install horizon-sdk[mcp] # Register with Claude Code export HORIZON_API_KEY="your-key" claude mcp add --transport stdio horizon -- python -m horizon.mcp ``` ## The Workflow ### 1. Morning scan - discover markets ``` You: Find me active bitcoin markets on Polymarket with recent activity ``` Claude calls `discover_markets(exchange="polymarket", query="bitcoin", limit=10)` and returns: ``` Found 6 active BTC markets on Polymarket: 1. will-btc-hit-100k - "Will BTC hit $100k by March 2026?" 2. will-btc-hit-120k - "Will BTC hit $120k by March 2026?" 3. btc-above-90k-feb - "Will BTC close above $90k in February?" 4. btc-dominance-60 - "Will BTC dominance exceed 60%?" 5. btc-etf-outflows - "Will BTC ETFs see net outflows this week?" 6. btc-halving-price - "BTC price on next halving anniversary" ``` ### 2. Analyze opportunity - Kelly sizing ``` You: I think there's a 72% chance BTC hits 100k by March. The market is at 58 cents. My bankroll is $5,000. How much should I buy? ``` Claude calls `kelly_sizing(prob=0.72, price=0.58, bankroll=5000.0, fraction=0.25, max_size=200.0)`: ``` Kelly analysis for your BTC $100k thesis: Edge: +14 cents (your 72% vs market's 58%) Quarter Kelly optimal: 200 contracts (capped at max_size) Breakdown: - Full Kelly fraction: 33.3% of bankroll - Quarter Kelly: 8.3% → $416 risked - At $0.58/contract: ~717 contracts uncapped - Capped at 200 contracts → $116 risk The edge is strong. Quarter Kelly at 200 contracts is conservative and well within risk limits. Want me to place the order? ``` ### 3. Place the trade ``` You: Yes, buy 200 contracts at 0.58 ``` Claude calls `submit_order(market_id="will-btc-hit-100k", side="buy", price=0.58, size=200.0)`: ``` Order submitted: Order ID: p1 Market: will-btc-hit-100k Side: buy Price: 0.58 Size: 200 contracts ``` ### 4. Add protection ``` You: Set a stop-loss at 0.40 and take-profit at 0.80 for that position ``` Claude calls `order_management(action="add_stop_loss", ...)` and `order_management(action="add_take_profit", ...)`: ``` Protection set: Stop-loss: Trigger: 0.40 (sell 200 YES contracts) Max loss if triggered: ~$36 Take-profit: Trigger: 0.80 (sell 200 YES contracts) Profit if triggered: ~$44 Your position is now bracketed. The stop-loss and take-profit will fire automatically when the market price hits those levels. ``` ### 5. Check the portfolio ``` You: Show me everything - positions, orders, pending stops ``` Claude calls `engine_status()`, `list_positions()`, `list_open_orders()`, and `order_management(action="list_contingent")`: ``` Engine Status: Running: yes | Kill switch: off | Uptime: 12 min Daily PnL: -$0.20 Positions (1): will-btc-hit-100k: 200 YES @ 0.58 Unrealized: -$0.20 | Exchange: paper Open Orders (0): All filled. Contingent Orders (2): 1. Stop-loss: sell 200 YES @ 0.40 (pending) 2. Take-profit: sell 200 YES @ 0.80 (pending) ``` ### 6. Second market - scan Kalshi ``` You: Are there similar BTC markets on Kalshi? Look for KXBTC ``` Claude calls `discover_markets(exchange="kalshi", query="KXBTC", limit=5)`: ``` Found 3 BTC markets on Kalshi: 1. KXBTC-26MAR31 - "BTC above $100k on March 31" 2. KXBTC-26FEB28 - "BTC above $95k on Feb 28" 3. KXBTC-26APR30 - "BTC above $110k on April 30" KXBTC-26MAR31 looks like a direct analog to your Polymarket position. If there's a price gap, that could be an arb. ``` ### 7. End of day - close out ``` You: Cancel all my open orders and show me final PnL ``` Claude calls `cancel_all_orders()` then `engine_status()`: ``` Canceled 0 orders (none were open). Final status: Realized PnL: $0.00 Unrealized PnL: +$4.00 Daily PnL: +$4.00 Your will-btc-hit-100k position is up $4 on 200 contracts. Stop-loss and take-profit are still armed. ``` ## Full Session as a Script For reproducibility, here's what the tool calls looked like: ```python theme={null} """Equivalent Python code for the MCP session above.""" from horizon import tools # 1. Discover markets = tools.discover("polymarket", "bitcoin", limit=10) # 2. Kelly sizing sizing = tools.kelly_sizing(prob=0.72, price=0.58, bankroll=5000.0) print(f"Optimal size: {sizing['optimal_size']} contracts") # 3. Trade order = tools.submit_order("will-btc-hit-100k", "buy", 0.58, 50.0) print(f"Order ID: {order['order_id']}") # 4. Protect tools.add_stop_loss("will-btc-hit-100k", "yes", "sell", 50.0, 0.40) tools.add_take_profit("will-btc-hit-100k", "yes", "sell", 50.0, 0.80) # 5. Check status = tools.engine_status() positions = tools.list_positions() contingent = tools.list_contingent_orders() # 6. Discover Kalshi kalshi_markets = tools.discover("kalshi", "KXBTC", limit=5) # 7. Close out tools.cancel_all_orders() final = tools.engine_status() print(f"Daily PnL: ${final['daily_pnl']:.2f}") ``` ## Next Steps Full tool and resource documentation. Detailed setup instructions. Deep dive into Kelly criterion functions. Full production deployment example.