# Amazon Bedrock adds structured JSON outputs for Claude > AWS announces schema-compliant JSON responses for Claude on both Bedrock endpoint paths. Engineers still need to validate the content. Canonical URL: https://www.devobs.io/news/news-event-53db7dc5301fd8a0d8fbc4fe/ By: Lena Fischer Published: 2026-10-21T15:35:07.471Z Updated: 2026-10-21T15:35:07.471Z Event date: 2026-10-20 Section: AI AWS announced structured-output support for all Anthropic Claude models on Amazon Bedrock on October 20, 2026. According to AWS, the feature returns JSON responses that conform to a defined structure, targeting applications that need predictable data rather than free-form answers. ([AWS announcement](https://aws.amazon.com/about-aws/whats-new/2026/10/claude-structured-outputs-bedrock/)) ## Support spans both Bedrock endpoint paths AWS says the feature covers both `bedrock-runtime` and `bedrock-mantle`. The announcement explicitly includes calls through boto3 and the Bedrock runtime APIs, alongside Messages API calls through the mantle endpoint. AWS describes the responses as schema-compliant JSON that does not require custom parsing, retry logic or post-processing. Its examples include document extraction, product catalog enrichment, contract analysis and support ticket routing. ([Endpoint coverage and use cases](https://aws.amazon.com/about-aws/whats-new/2026/10/claude-structured-outputs-bedrock/)) For engineers, the potential benefit is a simpler boundary between model output and application code: fewer format-repair steps before a response can enter a typed workflow. That is distinct from establishing whether the response contains the right information. ## Availability follows Claude model coverage AWS says structured outputs are available for all Anthropic Claude models on Bedrock, wherever those models are available. The announcement ties regional coverage to Claude model availability, so deployment planning should check the intended model and region. ([Availability statement](https://aws.amazon.com/about-aws/whats-new/2026/10/claude-structured-outputs-bedrock/)) ## Schema compliance is not answer verification Structural correctness and semantic accuracy are separate checks. A response can fit a required schema while naming the wrong contract party or assigning an unsuitable support category. Keep domain-specific validation, and use human review where the consequences warrant it. Likewise, AWS's statement about avoiding retry logic should not be a reason to remove handling for request failures: repairing malformed output and recovering from a failed request are different concerns. Treat structured outputs as a stronger interface contract—not as evidence that the underlying answer is correct. ## Source references -