Claude Managed Agents: Multiagent Orchestration with Specialist Subagents
Claude CodeClaude Managed Agents introduces multiagent orchestration as a first-class primitive, allowing a lead agent to delegate work to specialist subagents each configured with their own model, prompt, and tools. Subagents execute in parallel on a shared filesystem, with the lead agent able to check in mid-workflow as all events are persistent. Full traceability is available in Claude Console, showing which agent did what, in what order, and why. Netflix's platform team is already using this capability to run parallel batch analysis across hundreds of build logs, surfacing only actionable patterns from thousands of applications. The feature is now available in public beta.
What Is Multiagent Orchestration?
Claude Managed Agents now supports multiagent orchestration as a first-class primitive, enabling a lead agent to break complex tasks into discrete units and delegate each to a specialist subagent. Each specialist operates with its own model selection, prompt configuration, and tool access — making it possible to assemble heterogeneous agent teams that play to the strengths of different models.
How It Works
When a task arrives, the lead agent analyzes the work required and distributes subtasks to specialist subagents that execute in parallel on a shared filesystem. Because all agent events are persistent, the lead agent can check in with specialists mid-workflow, review intermediate outputs, and redirect work as needed. The subagents contribute their results directly back to the lead agent's overall context, allowing it to synthesize findings into a coherent final output.
Full traceability is available in Claude Console, where developers can inspect which agent performed which action, in what order, and why — providing the observability needed to debug, optimize, and audit agent behavior at scale.
Why It Matters
Multiagent orchestration transforms Claude Managed Agents from a single-agent assistant into a coordinated workforce capable of tackling tasks that would overwhelm any individual context window. Parallel execution dramatically reduces wall-clock time for large workloads, while specialist models ensure each subtask is handled by the most capable agent for that domain.
Netflix's platform team illustrates the real-world impact: their team built an analysis agent that processes logs from hundreds of builds across different sources. Multiagent orchestration enables parallel batch analysis to surface only the patterns worth acting on across thousands of affected applications — work that would be impractical with a single sequential agent.
Status and Availability
Multiagent orchestration is available in public beta. Developers can get started through Claude Console at platform.claude.com and access full documentation at platform.claude.com/docs/en/managed-agents/overview.