Claude Agent SDK v2.1.172 enables five-level subagent hierarchies
Anthropic expands agent-delegation architecture
Anthropic's Claude Agent SDK documentation now specifies that as of Claude Code v2.1.172, subagents can spawn their own subagents, enabling multi-tier task delegation workflows Claude Agent SDK. The update introduces a depth-limit safeguard: a subagent five levels below the main agent cannot spawn further subagents Claude Agent SDK.
This hierarchical delegation model allows a primary agent to assign work to a child agent, which can then create its own subordinate agents to parallelize subtasks. The architecture is relevant for complex workflows requiring multi-step problem decomposition—such as research coordination, financial analysis, or supply-chain optimization—where a single agent cannot efficiently manage all branches of execution.
Control mechanisms for spawn restrictions
Developers can prevent a subagent from spawning additional subagents using two methods: omit the `Agent` tool from the subagent's `tools` array, or explicitly add `Agent` to its `disallowedTools` configuration Claude Agent SDK. This granular control lets teams enforce different permission levels across the hierarchy—for example, allowing mid-tier agents to delegate but restricting leaf agents from further spawning.
The five-level maximum appears designed to balance flexibility against runaway delegation complexity. A typical workflow might use two or three levels; the cap prevents accidental or adversarial depth explosion that could degrade performance or create unpredictable control flow.
Developer adoption and use case implications
The feature ships as part of the Claude Code tooling ecosystem, positioning hierarchical agents for use in enterprise automation platforms, autonomous research systems, and multi-agent coordination frameworks. Teams building on Claude's infrastructure can now model organizational decision-making—where senior agents strategize, middle agents execute, and junior agents handle atomic tasks—within a single agentic workflow.
No official press release or announcement from Anthropic appears in public channels; the update surfaced through routine documentation maintenance. The change suggests incremental capability hardening rather than a flagship product launch, though it removes a significant architectural limitation in prior versions that did not support nested spawning.
The depth limit and tooling controls indicate Anthropic is prioritizing operational safety and predictability as developers integrate agents into production systems. Teams deploying hierarchical workflows should review their spawn-prevention policies to ensure subagents at critical decision points cannot inadvertently delegate beyond intended boundaries.