Claude Code v2.1.172: Sub-agents, Bedrock region, and numerous fixes
This patch for Claude Code introduces a significant enhancement allowing sub-agents to spawn their own sub-agents up to five levels deep, expanding agent capabilities. It also improves Amazon Bedrock integration by respecting AWS region configuration and includes fixes for session management, tool usage, and UI rendering. These changes benefit developers and architects using Claude Code for complex agent-based workflows, particularly those integrating with AWS services or encountering specific tool/UI bugs.
- →Sub-agents can now spawn their own sub-agents
- →Amazon Bedrock respects AWS region from config files
- →Plugin marketplace search bar added
- →OpenTelemetry metric includes model attribute
- →Fixed sessions with 1M context getting stuck
Features (1) ›
- Sub-agents can now spawn their own sub-agents
Agents can now create nested sub-agents up to five levels deep, enabling more complex and hierarchical agentic workflows within Claude Code.
Enhancements (3) ›
- Amazon Bedrock respects AWS region from config files
Amazon Bedrock integration now automatically reads the AWS region from `~/.aws` configuration files when `AWS_REGION` is not explicitly set, aligning with AWS SDK precedence. The `/status` endpoint now indicates the source of the detected region.
- Plugin marketplace search bar added
A new search bar has been added to the plugin browsing interface accessed via `/plugin`, making it easier to find specific plugins within a marketplace.
- OpenTelemetry metric includes model attribute
The `claude_code.lines_of_code.count` OpenTelemetry metric has been updated to include a `model` attribute, providing more granular data on model usage for code analysis.
Fixes (2) ›
- Fixed sessions with 1M context getting stuck
Sessions that previously used 1 million context without sufficient usage credits are no longer permanently stuck; they now automatically compact back under the standard context limit.
- Resolved repeating image processing errors
A recurring error message indicating that an image could not be processed and was removed has been fixed, particularly when multiple images were present in a conversation.
https://github.com/anthropics/claude-code/releases/tag/v2.1.172