In the future, your agents scroll and trade for you.
Clone the repo, install dependencies, then run npm run agent.
git clone https://github.com/VittorioC13/Musashi.gitcd Musashi && npm installnpm run agent
A cleaner quick start for customer onboarding.
Monitors 71 high-signal Twitter accounts across 8 categories and matches tweets to markets every 2 minutes.
AI-powered sentiment analysis classifies each tweet as bullish, bearish, or neutral with confidence scores, direction, edge, and urgency.
Cross-platform price discrepancy detection between Polymarket and Kalshi with filters for spread, confidence, and category.
Tracks major price changes using stored price history and real-time polling to surface 1 hour and 24 hour movement.
Everything you need to know about Musashi
Andrej Karpathy said in his tweet: "it's 2026, Build. For. Agents" - and that's why we built Musashi. We believe the era of CLI and agent-first products is here. Humans shouldn't be glued to screens watching markets - agents should monitor, analyze, and trade autonomously. Musashi provides the structured intelligence layer that agents need: not a pretty dashboard, but a high-frequency API with trading signals, confidence scores, and real-time market data. We're building infrastructure for the autonomous trading era.
Musashi is an AI intelligence service that automatically collects and analyzes tweets from 71 high-signal accounts every 2 minutes, matches them to 900+ prediction markets, generates trading signals, and feeds structured data to AI trading bots via a polling API. The Chrome extension is just a monitoring dashboard (10% of the product) - the real product is the intelligence layer for automated trading.
Primarily for AI trading bots and developers building automated prediction market strategies. The Chrome extension lets humans monitor what the bots see, but the core product is the feed API that provides analyzed tweets with sentiment, urgency levels, confidence scores, and suggested trading actions.
Every 2 minutes, Musashi collects tweets from 71 monitored accounts across 8 categories (crypto, politics, economics, tech, sports, geopolitics, finance, breaking news). Each tweet is analyzed using keyword matching, entity extraction, and sentiment analysis, then matched to relevant markets. Only tweets that match ≥1 market are stored in the feed for 48 hours.
For each analyzed tweet: original author and text, 1-5 matched markets with confidence scores, sentiment (bullish/bearish/neutral), suggested action (YES/NO/HOLD with confidence + edge), urgency level (critical/high/medium/low), category, and reasoning. Everything your bot needs to make trading decisions.
Use the Agent SDK (TypeScript/JavaScript) or call the REST API directly. Poll /api/feed every 20-30 seconds with filters (category, urgency, since timestamp) for optimal freshness. The SDK provides callback methods like onFeed() that handle polling automatically. Python, Node.js, and other language examples are in the docs.
900+ markets from Polymarket (500) and Kalshi (400), sorted by volume. Politics, economics, crypto, technology, sports, geopolitics, climate, and entertainment. Markets update every 15-20 seconds for real-time trading decisions. Coverage includes all major markets across categories with fresh pricing data.
Yes, completely free during beta. The feed API, Chrome extension, and SDK are all free with no rate limits currently. All infrastructure runs on free tiers (Vercel, Vercel KV, Twitter API). Fair usage limits may be added in the future to ensure service reliability, but the core service will remain free for reasonable bot usage.
Musashi detects cross-platform arbitrage opportunities (e.g., same market trading at 63% on Polymarket, 70% on Kalshi = 7% spread) and tracks markets with significant price changes (>5% in 1h or 24h windows). Access via /api/markets/arbitrage and /api/markets/movers endpoints.