A Guth Labs publication ↗
Back to News

Remanence raises $30 million to train models on AI agent traces

Green glass AI core surrounded by interchangeable metal modules.
AI-generated editorial illustration

Remanence raised $30 million in seed funding to help companies train AI models on the execution records generated by their agents, according to EQT Ventures' November 17 announcement. EQT Ventures and FirstMark co-led the round.

The Paris-based startup turns agent traces, including tool calls, failures and corrections, into training environments and evaluation mechanisms. Those environments support reinforcement learning on open-weight models tailored to a customer's work. EQT says customers own the resulting model weights and can keep their data within their own infrastructure.

Participating investor Zero Prime says the platform can run as a managed service or in a customer's cloud. Other investors include Emblem, Drysdale, Global Founders Capital and Yellow.

Founders Pierre-Louis Cedoz, Hamza Benchekroun and Axel Moyal previously worked together on H Company's Holo models. The funding will support a dedicated GPU cluster in Europe, research and engineering hiring, and expanded design partnerships in Europe and the United States.

The approach offers companies a way to reuse their agents' operational experience rather than rely solely on general-purpose models. The announcement does not establish independently measured performance or cost improvements.

Sources and citations

Revision history

Published .