How to Build a Prediction Market Trading Bot with Musashi

Prerequisites

You need Node.js 18+ and a Polymarket or Kalshi account to execute trades. The Musashi API requires no authentication and has no rate limits during beta.

• →Node.js 18+
• →Git
• →Polymarket or Kalshi account (for execution)
• →Optional: Python 3.9+ for Python bots

Step 1: Install Musashi

git clone https://github.com/MusashiBot/Musashi.git
cd Musashi && npm install
npm run agent

The agent starts polling Twitter accounts and building the market match cache. The first full cycle takes about 2 minutes.

Step 2: Poll the Feed API

Your bot polls GET /api/feed to get new signals. Filter by category and urgency to reduce noise.

const BASE = 'https://musashi-api.vercel.app'
let lastSeen = Date.now()

async function pollFeed() {
  const res = await fetch(
    `${BASE}/api/feed?urgency=high&since=${lastSeen}`
  )
  const { signals } = await res.json()
  const newestSignalTs = signals.reduce(
    (max, signal) => Math.max(max, signal.createdAt),
    lastSeen
  )
  lastSeen = newestSignalTs
  return signals
}

setInterval(async () => {
  const signals = await pollFeed()
  for (const signal of signals) {
    if (signal.sentiment === 'bullish' && signal.confidence > 0.75) {
      await executeTrade(signal)
    }
  }
}, 25_000)

Step 3: Interpret Trading Signals

Each feed item contains everything your bot needs to make a trading decision:

Step 4: Execute Trades

Use the Polymarket CLOB client or Kalshi API to place orders when your signal threshold is met. Only trade markets where signal confidence is above your minimum threshold (0.70+ recommended to start).

For arbitrage opportunities, check GET /api/markets/arbitrage — see the full reference in the API docs.