TL;DR: Anthropic released Claude Code 2.1.211, enabling subagent thought processes to be streamed in JSON for better downstream system integration.
Summary: Anthropic has released Claude Code 2.1.211, featuring 37 CLI changes. A key highlight is the addition of a flag/environment variable to include subagent text and thinking in stream-json output. This allows downstream systems to receive and process the subagent's internal reasoning.
Why it matters: AI developers can now access and utilize the intermediate thought processes of Claude's subagents, enabling more sophisticated agentic workflows and debugging. Experiment with the new streaming capabilities to build more transparent and controllable AI applications.
Source: x_com