AWS adds schema-constrained Claude responses to both Bedrock endpoints

AWS says structured outputs are now generally available for all Anthropic Claude models wherever they are available on Amazon Bedrock's bedrock-runtime and bedrock-mantle endpoints. The feature constrains generated tokens to produce JSON matching a supported schema, helping developers extract data and pass structured information to applications. It supports batch inference and streaming responses, plus cross-Region inference on bedrock-runtime. Strict tool use on bedrock-mantle constrains tool arguments to their declared schemas. The guarantees have limits: refusals and token limits can produce nonconforming responses. Numerical bounds and string-length constraints are not enforced during generation and require application validation. New schemas also incur compilation overhead; compiled grammars are cached for 24 hours.