AWS has outlined how regulated organizations can connect Claude Code to Claude 5.5 models through Amazon Bedrock in GovCloud.
AWS has published a deployment path for connecting Claude Code to Claude Opus 5.5 and Claude Sonnet 5.5 through Amazon Bedrock in AWS GovCloud (US). The guide targets US organizations operating regulated workloads and covers model access, endpoint selection, centralized controls and limitations that should be assessed before an enterprise rollout.
Claude Code is Anthropic’s agentic development tool. According to AWS, it can read a codebase, edit files, execute commands and tests, inspect Git history, and create commits and pull requests. It operates in terminals, VS Code and JetBrains environments, or background processes through the Claude Agent SDK. External tools and data sources can be connected through the Model Context Protocol.
Deployment requires an AWS GovCloud (US) account with Amazon Bedrock access, appropriate IAM permissions, enabled models and an AWS CLI session using valid credentials. Administrators can use Claude Code’s interactive setup wizard or configure environment variables. AWS advises pinning full model identifiers because built-in aliases may change between releases; Claude Opus 5.5 is the tool’s default primary model.
Bedrock offers two endpoint surfaces. The bedrock-runtime endpoint is available in GovCloud US-West and US-East and supports AWS SDK APIs, Guardrails, Knowledge Bases, Agents and invocation logging. The bedrock-mantle endpoint is limited to GovCloud US-West but natively supports the Anthropic Messages API, server-side tools, background inference and Projects. AWS says Guardrails and invocation logging are exclusive to bedrock-runtime.
Model-level authorizations differ. AWS lists Claude Opus 5.5 and Claude Sonnet 5.5 as holding FedRAMP Class D certification on Amazon Bedrock. For workloads requiring DoD Impact Level 4 or 5 authorization, it recommends Claude Sonnet 5. Organizations are directed to verify current certification status because compliance depends on the selected model and their applicable obligations.
The architecture is a basis for an organization’s own review, not an automatic compliance guarantee. Bedrock secures the inference layer, but Claude Code runs on developers’ local machines and can execute commands or modify files. That separation makes managed permissions, temporary role-based credentials, invocation logging, token limits and centrally administered settings important controls before broader deployment.
AWS warns that the approach may not suit every organization or compliance program and calls for a thorough security assessment. It also recommends reviewing request and token quotas, monitoring costs and, where appropriate, enforcing daily per-developer token limits. The post identifies Sonnet 5.5 as less expensive than Opus 5.5, but provides no specific prices or workload cost estimates.
| Endpoint | Regions | Key capabilities | Limitation |
|---|---|---|---|
| bedrock-runtime | US-West and US-East | AWS SDK, Guardrails, Knowledge Bases, Agents and invocation logging | No native Anthropic Messages API surface |
| bedrock-mantle | US-West | Anthropic Messages API, server-side tools, background inference and Projects | No Guardrails or invocation logging |
Sources
Originally published by AI Feed: AWS Details Claude Code Deployment in GovCloud with Claude 5.5.
Top comments (0)