Antithesis Adds AI Agent Tools to Its Deterministic Testing Platform

Antithesis has added AI coding-agent support to its deterministic debugging platform, letting agents replay software failures and test fixes in controlled simulations, while extending a tool previously used in Ethereum testing.

Announced Thursday, Oct. 15, the update makes Antithesis’ Multiverse Debugger available to AI coding agents. They can replay a failed run, inspect the system’s history and state, issue commands in a deterministic environment and try a code change under the same conditions. Agents can also rewind execution and branch from a point in a run, allowing proposed fixes to be checked against the failure state.

Antithesis’ testing platform runs software in automated simulations, injects faults and explores edge cases. Making behavior repeatable helps developers reproduce failures and check whether a change resolves them. The agent tools bring those debugging actions into workflows where AI systems generate code.

“Agents can author code faster than we can check it,” said David Koenitzer, Antithesis’ product manager for Time Travel. The update gives agents tools to inspect an execution, reproduce a failure and test a correction inside the simulation.

Jane Street led Antithesis’ $105 million Series A, announced Dec. 3, 2025. The trading firm was also a customer. The financing was intended to support engineering, platform intelligence and autonomy, sales and marketing, and expansion across North America, Europe and Asia. Antithesis also planned to make the platform available through cloud channels, including AWS Marketplace.

Jane Street used Antithesis to test its Aria distributed system. Engineer Doug Patti wrote, “Sure enough, on our first run, Antithesis surfaced two previously unknown bugs.”

Antithesis has also been used to test Ethereum. The Ethereum Foundation included coverage-guided network fuzzing using Antithesis in its March 23, 2022, Merge testing update. Before the upgrade, Antithesis ran Ethereum network code in simulation to find and reproduce failures.