Anthropic Introduces Advisor Feature for Claude Managed Agents
Anthropic has added a new feature to Claude Managed Agents that allows a working agent to consult a stronger "advisor" model mid-session for a second opinion. This hierarchical pattern can be enabled with just a single line of configuration in the agent's roster. This update simplifies the implementation of hierarchical agent architectures, enabling developers to build more cost-effective systems where cheaper models handle routine tasks and escalate to advanced models only when necessary. The feature leverages Anthropic's advisor strategy, typically pairing a cost-efficient executor model like Claude Sonnet or Haiku with a more powerful model like Claude Opus for complex decision-making.
## BACKGROUND
Claude Managed Agents is a service on the Claude Platform that provides the pre-built, configurable harness and managed infrastructure needed to run autonomous AI agents at scale. The advisor strategy is a design pattern aimed at balancing performance and API costs by using a hierarchical multi-agent setup.