The problem is not “a second Polymarket UI.” The problem is Kalshi’s election/event books
plus a research note you can distrust. KalshiEdge talks to
https://api.elections.kalshi.com with RSA-PSS signed requests:
open markets (yes/no + volume), market detail, trades, your positions, balance, fills.
For one ticker it runs two DuckDuckGo searches then OpenRouter
anthropic/claude-haiku-4-5 and caches a structured briefing 6 hours in
kalshi_claude_analyses: facts, movers, gaps, similar markets, YES/NO/UNCERTAIN +
confidence + one-sentence rationale. Watchlists and price alerts are per-user. Flask-Login, not
SWIP. Public landing at /apps/kalshi/ — /kalshi/ 404s; that is the
prefix, not a missing app. Phase 0 calibration is SHELVED (negative result).
This is not a broker, not auto-trade, not financial advice.
If KALSHI_KEY_ID or the PEM is missing, the books are empty — that is config, not a chart bug.
Not Kalshi.com chrome. Not Polymarket. Not a fake P&L screenshot.
YES/NO/UNCERTAIN + HIGH/MEDIUM/LOW. 6h kalshi_claude_analyses.
Negative empirical result. Do not advertise a calibrated edge that was killed.
Markets move. This VPS is not your introducing broker.
Will not place orders. Landing lists no fills. OpenRouter is the analysis path by design.
Operator with KEY_ID + PEM. If either is missing, the storefront still tells the truth.
Cards match the Phase 1 README.
Login for the app. This page is not your portfolio.
Open KalshiEdgeRequires a MelTuc account. Create one free.
Prediction-market decision support — browse Kalshi markets and get analysis briefs. Health stays cheap. If this page disagrees with the signed-in app, trust the app — the product contract and tests enforce that rule.