The langchain-anthropic integration has been updated to version 1.7.2, a focused patch release that addresses a specific issue with how tool use blocks are handled.
Previously, when the Anthropic model returned tool use blocks that were considered invalid, the integration would discard them. With this update, those blocks are now preserved instead of being dropped. For self-hosted users working with Anthropic models, this means malformed or unexpected tool calls will no longer disappear silently. Instead, they are passed through so you can inspect them, log them, or apply your own error handling — giving you better visibility and control over model behavior.
