The Agent Memory Challenge 2026 Cycle 2 has opened to evaluate long-term memory capabilities of AI agents, emphasizing retrieval of useful evidence amid changing facts and tasks. Participants can submit textual, coding, or multimodal memory approaches, including both open‑source methods and commercial products. Details and registration are available via the challenge evaluation page.