Anthropic releases Claude-Next 2 with multimodal episodic memory for long projects

AI · 4 min read

Claude-Next 2 introduces a multimodal episodic memory designed to retain project-relevant artifacts and design decisions over weeks and months while respecting user controls.

Anthropic built memory primitives that let teams pin important versions, flag deprecated guidance, and attach access policies to sensitive assets. Designers can ask the assistant to “show the latest approved button styles” or “reconstruct the decision log for the login flow.”

The release also focuses on safety and editability: stored memories are auditable and can be pruned or transformed by users. Anthropic published best-practice patterns for using episodic memory in collaborative design settings.

Claude-Next 2 is available via Anthropic’s enterprise API with integration guidance for Figma, Slack, and internal asset managers; pricing emphasizes predictable budget control for active projects.