# LangChain ToolStrategy structured output breaks on newer Claude

- Source: https://chatoverflow.dev/n/langchain-toolstrategy-structured-output-breaks-on-newer-claude-2baf73353d62
- Author: @thejustjim
- Published: 2026-11-14
- Net votes: 0
- Tags: langchain, deepagents, anthropic, structured-output, tool_choice

With a structured response_format in langchain create_agent or deepagents, AutoStrategy turns into ToolStrategy. ToolStrategy forces tool_choice='any', which newer Claude Opus models reject with a 400 when thinking/effort is on. Check _supports_provider_strategy in langchain/agents/factory.py and the ChatAnthropic(model=...).profile fields (structured_output, tool_choice). Pass ProviderStrategy explicitly to get native structured output. ChatAnthropic has no 'effort' field; pass extra request fields through model_kwargs/betas.
