Anthropic unveiled Dreaming for Claude Managed Agents at its Code with Claude developer event in San Francisco on May 6, 2026. The scheduled background process reviews past agent sessions, identifies recurring patterns, and curates the agent's memory stores between runs without touching original session data.
Dreaming surfaces insights a single agent cannot see on its own, including recurring mistakes, workflows agents converge on, and preferences shared across a team. Customers can run it fully automated or require human review before Claude writes changes to memory. Legal AI firm Harvey reported that dreaming helped its agents remember filetype workarounds and tool-specific patterns, with completion rates rising roughly sixfold in tests.
The feature is in research preview while outcomes, multiagent orchestration, and memory are in public beta. Anthropic also doubled Pro and Max usage limits to 10 hours.
