Claude Managed Agents Can Engage In a 'Dreaming' Process To Preserve Memories

Source: Slashdot

Dreaming for Claude Managed Agents

At its Code with Claude developers’ conference, Anthropic introduced a feature called “dreaming” for Claude Managed Agents. Dreaming is a process that reviews recent events and identifies specific items worth storing in memory to inform future tasks and interactions. This feature is currently in research preview and limited to Managed Agents on the Claude Platform.

Managed Agents

Managed Agents are a higher‑level alternative to building directly on the Messages API. Anthropic describes them as a “pre‑built, configurable agent harness that runs in managed infrastructure.” They are intended for scenarios where multiple agents collaborate on a task or project over several minutes or hours.

How Dreaming Works

• Dreaming is a scheduled process that reviews sessions and memory stores, curating specific memories.
• Because LLM context windows are limited, important information can be lost during lengthy projects.
• Dreaming periodically analyzes past sessions and memory stores across agents, identifying important patterns and saving them for future use.
• Users can choose between an automatic dreaming process or manually reviewing changes to memory.

Relation to Conversation Compaction

On the chat side, many models use compaction, which periodically analyzes lengthy conversations to remove irrelevant information from the context window while retaining what’s essential for the ongoing dialogue. Compaction is typically limited to a single conversation with a single agent. In contrast, dreaming operates across agents and sessions, providing a broader, recurring memory‑curation mechanism.