SKALE Labs launched Agent Pit on Aug. 12, a paper-trading sandbox that lets AI agents test prediction market strategies without touching live funds.

Agent Pit runs on SKALE's gas-free chain and replicates Polymarket's order book, market settlement structure and real-time event data feeds.

The goal is to give developers confidence in an AI agent's mechanics and strategy before going live, SKALE Labs CEO Jack O'Holleran said. Agents can then carry refined strategies into live markets such as Polymarket.

AI agents on Polymarket executed more than 4,200 trades in the month before Agent Pit's announcement, generating returns in live market conditions.

SKALE's V4 upgrade, which went live in Jan. 2026, was built with AI agent applications in mind. It introduced infrastructure improvements, privacy tools and agent payment rails that allow bots to transact without constant human intervention.

The Ethereum-compatible network is positioning itself as infrastructure for autonomous AI trading, starting with prediction markets—whose binary or categorical outcomes and deterministic resolution make them well-suited for AI systems.

Agent Pit includes performance leaderboards that benchmark strategies and add a competitive element to agent development.

The platform's value depends on how accurately it captures live Polymarket conditions—including liquidity dynamics, slippage and adversarial behavior. SKALE says Agent Pit mirrors Polymarket's mechanics precisely, a claim that will be tested as agents trained in the sandbox begin deploying real capital.