Product Overview
Prexora is one engine exposed through several surfaces. The engine finds and bets edge on weather markets; the surfaces let you watch it, follow it, copy it, and delegate to it.
The one engine
At the center is an autonomous agent that runs a continuous loop: ingest a forecast ensemble and market prices, model true probabilities, find edge, gate and size, place the bet, resolve it, and reflect. Everything else is a window onto that loop or a way to act on its output. There is no separate “demo” engine and “real” engine — every surface reads the same live state.
The surfaces
Live Signals Board
Today’s open weather brackets where the agent has edge, with strength tiers and rationale. Live
How it works →Mission Control
A real-time view of the agent’s decision flow — what it’s scanning, deciding and booking. Live
The Engine →Track Record
Verifiable performance on funded signals only, anchored on-chain. Live
Transparency →Lesson Stream
The agent’s written reflection on every resolved bet, wins and losses. Live
View →Telegram
New picks and wins pushed to @prexora in real time. Live
Join →Copy & Autonomous
One-click copy a pick, or fund a delegated account and let the agent trade. Live
Get started →Three ways to use it
| Mode | You do… | The agent does… |
|---|---|---|
| Manual | Read the board and place your own trades on Polymarket. | Finds and publishes the edge. |
| One-click copy Live | Tap to mirror a pick into your own order. | Finds the edge; you keep custody and control. |
| Autonomous Live | Fund a delegated account once. | Finds, sizes, places and manages bets for you under the same risk guards. |
What makes it different
- A verifiable process — signed signals and a daily on-chain anchor, not screenshots.
- A weather niche that is physically grounded, decorrelated from crypto, and underserved by quant players.
- Consensus, not a single guess — independent layers (the resolver-aligned WU forecast, a 3-model ensemble, real-time METAR, a 12-agent AI swarm, and whale/leaderboard signals) must agree before real money is risked.
- Honest breadth — display vs funded separation lets the board show 20–40 daily signals without inflating performance.
- A multi-guard risk system that defends capital from every angle — city, event, day and open trade.
- A learning loop that compounds the edge, with every change graded before it touches money.
- Self-healing operations so the agent keeps trading through API hiccups without manual babysitting.