The Pulse
AWS Details Claude Code Setup for Claude 5.5 in GovCloud
AWS published instructions for running Claude Code with Claude Opus 5.5 and Claude Sonnet 5.5 through Amazon Bedrock in AWS GovCloud. The models carry FedRAMP Class D certification, but AWS identifies Claude Sonnet 5—not the 5.5 models—as a
AI.info Team · Written with AI ·
AWS is opening a path for Claude Code in regulated GovCloud environments while drawing a clear line around which models carry which approvals. In a technical post published October 5, AWS describes using Claude Opus 5.5 and Claude Sonnet 5.5 through Amazon Bedrock in AWS GovCloud, but says the models’ FedRAMP Class D certification does not give them the DoD IL4 and IL5 authorization held by Claude Sonnet 5.
FedRAMP certification is not DoD authorization
The distinction matters for agencies and contractors choosing models for sensitive work. AWS says both Claude 5.5 models have FedRAMP Class D certification on Bedrock; it identifies Claude Sonnet 5 as the model with DoD Impact Level 4 and 5 authorization. AWS recommends Sonnet 5 for workloads that require IL4 or IL5, rather than treating the newer models’ presence in GovCloud as equivalent clearance.
AWS had announced Claude Opus 5.5 availability in GovCloud on September 22. The October 5 post focuses on connecting Claude Code to the models and describes Claude Sonnet 5.5 alongside Opus 5.5. AWS frames the setup as an option for regulated workloads, including those subject to the International Traffic in Arms Regulations, while advising organizations to assess it against their own compliance obligations.
Two Bedrock endpoints, different controls
The setup routes Claude Code through one of two Bedrock endpoints. The bedrock-runtime endpoint is available in both GovCloud regions, US-East and US-West, and supports Guardrails, invocation logging, Knowledge Bases, and Agents. AWS recommends it for deployments that need audit trails or content filtering.
Bedrock Mantle, available in GovCloud US-West, supports the Anthropic Messages API natively and offers features such as server-side tools, background inference, and Projects. But Guardrails and invocation logging are exclusive to bedrock-runtime. Organizations choosing Mantle therefore give up those particular Bedrock controls in this configuration.
Claude Code needs local security review
Claude Code can read and edit a codebase, run commands, execute tests, and connect to tools such as the AWS Command Line Interface, Terraform, and Kubernetes. Developers can configure it through an interactive login wizard or environment variables; AWS gives us-gov-west-1 as the region and recommends pinning a full model ID for consistent deployments.
That setup does not move all security responsibility into Bedrock. AWS says Claude Code runs on developers’ local machines and needs its own security assessment. The post advises organizations to manage access with IAM Identity Center, use temporary role-based credentials, review permissions and connected tools, and monitor token use.
Model selection affects cost and access
AWS warns that Claude Code defaults to Opus 5.5 as its primary model, so teams that leave the model unpinned may be billed at Opus rates. The company recommends setting the model explicitly—such as Sonnet 5.5 for lower-cost use—and using Opus for tasks that call for deeper reasoning or longer autonomous runs.
The practical dividing line is specific: AWS describes a way to use Claude Code with both 5.5 models in GovCloud, but directs workloads requiring DoD IL4 or IL5 authorization to Sonnet 5. The runtime endpoint offers Guardrails and invocation logs in both GovCloud regions; Mantle’s native Messages API is available only in US-West.