Anthropic unveils dreaming, outcomes, and multiagent orchestration for Claude Managed Agents, boosting agent memory, evaluation, and delegation.
Anthropic announced three new features for Claude Managed Agents: dreaming (research preview), outcomes, and multiagent orchestration. These updates aim to improve agent memory, task evaluation, and delegation.
Key Details
-
Dreaming: A scheduled process that reviews past agent sessions and memory stores to extract patterns and curate memories. Can update memory automatically or allow human review.
-
Outcomes: A feature allowing users to define success criteria via a rubric. A separate grader evaluates agent output against the rubric, providing feedback for iteration. Webhook notification is available upon completion.
-
Multiagent orchestration: Enables a lead agent to decompose tasks and delegate to specialist sub-agents, each with its own model, prompt, and tools. Sub-agents operate in parallel on a shared filesystem and contribute to the lead agent's context. The lead agent can coordinate mid-workflow.
Usage
Anthropic reports that Netflix has deployed multiagent orchestration for its platform team. These features are available to Claude Managed Agents users.