Agents
forecast.
Prophet
executes.
API and interface that turns your agent's beliefs into the best real orders across prediction markets. What, where and how to trade.
Separate order books to one interface.
The same markets quote differently on each venue. Prophet reads each book, merges them, and routes your order where it fills best.
The bottleneck to agentic trading on prediction markets is trade execution.
How do we get from agent-forecasted odds to actual trades?
If prediction market trading is going to be mostly agentic, developers will need one reliable interface across venues to translate beliefs and instructions into the most optimal real orders that can be matched.
LLMs are excellent at pricing on prediction markets
2024
LLMs used for forecasting real-world events; trail humans
Halawi et al., NeurIPS 20242025
AI bot places 8th in Metaculus Cup
2026
AI bot wins Metaculus Cup (1st of 885)
Today
LLMs match “superforecasters”
ForecastBench, Jul 2026
Using your agent's forecast, Prophet tells it what, where and how to trade.
Find unified data
Normalized markets, books and resolution rules across venues.
Match event across venues
The same event is identified and reconciled across contracts.
Compare price, liquidity, fees, rewards
Expected returns and costs to you.
Route best orders to best market and venue
Multiple low-latency orders with confirmation returned.
- 01“I think the weather in NYC tomorrow is going to be warmer than today.”
- 02“Make a trade on gold prices going up tomorrow compared to today.”
- 03“I want to trade on Gavin Newsom becoming the next president.”
- 04“Make 100 orders on Kalshi’s Oscars Best Picture market within 2¢ of the midpoint every 5 minutes for 30 minutes.”
- 05“I believe the odds of Ground Zero winning the CS2 game today are 80%.”