Vibe-Trading Setup Tutorial: Deploy Your AI Trading Agent in 15 Minutes

📝 Setting Up Your First AI Trading Agent in 15 Minutes

I spent four hours on my first setup because I kept missing little things. This guide exists so you don't make the same mistakes. Follow it step by step and you'll have a working trading agent in 15 minutes flat.

What you'll need: Docker (installed and running), about 15 minutes, and maybe a coffee.

⚡ Quick Start: Docker (The Easy Way)

Step 1: Pull and Run

docker pull akashtrade101/vibe-trading:latest
docker run -d \
  --name vibe-trading \
  -p 8080:8080 \
  -v ~/.vibe-trading:/root/.vibe-trading \
  -e LLM_PROVIDER=openai \
  -e LLM_API_KEY=sk-your-key-here \
  akashtrade101/vibe-trading:latest

If you see Started server on 0.0.0.0:8080 in the logs, you're golden. If not — check your Docker daemon. This is the step that tripped me up when my Docker wasn't running.

Step 2: Open the Web UI

Open your browser and go to http://localhost:8080. You should see the Vibe-Trading dashboard with a chat interface, portfolio view, and settings panel.

🧪 Testing with the Shadow Account (Don't Skip This)

I made the mistake of connecting a real brokerage account on day one. Don't be me.

The Shadow Account is a fully simulated trading environment with virtual money. It connects to real market data but executes trades in a sandbox. Think of it as a flight simulator for trading.

# Start a shadow account session
vibe-trading shadow start --balance 50000

# Check performance
vibe-trading shadow status

# Run a backtest on your strategy
vibe-trading backtest run --strategy mean-reversion --symbols SPY,AAPL,TSLA

I started with $50,000 of virtual money and a simple momentum strategy. After 3 days, I was down 2%. The agent logged every trade, and I could see exactly where it went wrong — bought the top of a pump. The self-improvement engine kicked in and adjusted the entry conditions. By day 7, I was up 4%.

🔌 Connecting Claude Code / Codex via MCP

This is where Vibe-Trading becomes incredible. Its built-in MCP (Model Context Protocol) server lets you connect AI coding agents directly to your trading environment.

# In your Claude Code config
{
  "mcpServers": {
    "vibe-trading": {
      "command": "python",
      "args": ["-m", "vibe_trading.mcp"],
      "env": {
        "VIBE_TRADING_URL": "http://localhost:8080"
      }
    }
  }
}

Now your AI coding agent can run backtests, check portfolio status, and even place trades (with the mandate gate enabled, of course). I told Claude Code "analyze my shadow account performance and suggest strategy improvements" — it ran a backtest, checked the correlation matrix, and suggested switching from momentum to mean-reversion for the current market regime.

🏗️ Docker Compose (Production Setup)

For a proper deployment, use Docker Compose with all the services. Here's what the generated compose file on Run This Ai sets up:

services:
  vibe-trading:
    image: akashtrade101/vibe-trading:latest
    ports:
      - "8080:8080"
    volumes:
      - ~/.vibe-trading:/root/.vibe-trading
    environment:
      - LLM_PROVIDER=openai
      - LLM_API_KEY=${LLM_API_KEY}
      - DATA_PROVIDER=yfinance
    restart: unless-stopped

❌ Common Mistakes I Made

🎯 What's Next?

Once you've got the basic setup running, try these:

• Set up a multi-agent team — one for technical analysis, one for fundamentals, one for risk
• Explore the Alpha Zoo — run vibe-trading alpha list to see 460+ factors
• Connect IM channels — get trading alerts on Telegram or Discord
• Run scheduled research — the agent can analyze markets on a daily schedule and deliver reports to your channels

The great thing about Vibe-Trading being open-source (MIT) is that you can dig into the code, customize strategies, and contribute back. The community is active on GitHub with 24K+ stars and growing.