Claude Opus 5.5, Sonnet 5.5 land on Amazon Bedrock in AWS GovCloud (US)

The arrival of Claude Opus 5.5 and Claude Sonnet 5.5 on Amazon Bedrock in AWS GovCloud (US) is framed by the article as an on-ramp for AI-assisted development where regulatory or compliance requirements apply. The two Claude 5.5 models each carry FedRAMP Class D certification on Amazon Bedrock, while Claude Sonnet 5, already in the catalog, holds both FedRAMP Class D (formerly High) certification and DoD IL4/IL5 authorization. That difference in compliance coverage is why the article suggests teams defaulting to Claude Sonnet 5 for workloads requiring IL4/IL5, and treating Claude Sonnet 5.5 as a step up on coding and knowledge work at a lower cost per task, with Claude Opus 5.5 reserved for deeper reasoning and longer autonomous runs.

Operationally, the post walks through installing Claude Code and pointing it at Bedrock using either the interactive setup wizard, manual environment variables, or the Bedrock Mantle endpoint. It places weight on separating the inference layer, which Amazon Bedrock secures, from Claude Code itself, which runs on local developer machines and needs its own security review. The article also points to IAM Identity Center, managed permissions, CLAUDE.md memory files, and per-user token guardrails as controls that govern how the tool is rolled out, alongside caveats that Guardrails and invocation logging are exclusive to the bedrock-runtime endpoint.

What this hinges on is whether an individual organization's compliance program treats this configuration as suitable; the article explicitly says the approach may not fit every organization and that current model certification status should be verified. For security and platform teams in regulated environments, the practical test is likely whether their own review, quota planning, and endpoint choice line up with the authorization pathway they need.