Amazon Bedrock supports structured outputs for all Claude models

Amazon Bedrock now provides structured outputs for all Anthropic Claude models across the bedrock-runtime and bedrock-mantle endpoints. The feature uses constrained decoding to enforce JSON Schema during token generation, making output format deterministic rather than probabilistic. Developers pass a schema via output_config.format, or set strict: true on tool definitions to strictly validate tool parameters. JSON Schema support covers a subset of Draft 2020-12; compiled schemas are cached per account for 24 hours, so the first request with a new schema is slightly slower while later requests incur minimal overhead. Developers using the Claude API get valid, type-safe JSON directly, eliminating JSON.parse() exception handling and retries for schema violations. Structured outputs also constrain the model to generate only schema-defined fields, reducing output tokens and lowering inference cost for high-concurrency batch workloads. The feature supports batch inference, cross-region inference, and streaming structured responses. For teams building agents on Amazon Bedrock Mantle, strict tool use ensures tool parameters precisely match the input schema, preventing multi-step workflows from failing due to malformed parameters. Use cases like data extraction, ticket classification, and contract analysis can connect directly to databases or CRMs without extra validation code.