Observed arrival · 2026-08-31
tradefloor makes AI trading agents face a reproducible market
A market simulator lets researchers evaluate financial AI agents against seeded prices, order-book execution, macro scenarios, and recorded factor attribution.
Field notes
The simulator builds a market from an explicit universe, macro state, and seed, then advances prices and an order book as an agent trades. Its run manifests preserve the strategy, model, scenario, and fingerprints, while reproduce() checks whether a later result matches the recorded digest. The example runs 30 days and reports measures such as spread, mispricing, VIX, and return spread. Documentation also describes Python, Gymnasium, and MCP integration paths.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue