ApplicationsBusiness & Market 🇺🇸 06.08.2026 20:02

Enforcing Data Residency with Single-Region Claude Code on Amazon Bedrock

AnthropicAnthropic Amazon Web ServicesAmazon Web Services
A US global organization required Amazon Bedrock model inference for Claude Code to be processed in London only. AWS explored Anthropic's Mantle endpoint and the classic Bedrock Invoke API with an application inference profile. The latter, combined with an IAM Region condition, satisfied the requirement. The post details both paths, the IAM policies, and verification via AWS CloudTrail.
A US-headquartered global organization asked AWS to let their engineers use Claude Code while ensuring all Amazon Bedrock model inference (prompts, completions, intermediate processing) stays within a single AWS Region, specifically London (eu-west-2). AWS first tried Anthropic's Mantle endpoint on Amazon Bedrock but hit a limitation, so they fell back to the classic Amazon Bedrock Invoke API with an application inference profile, plus an IAM Region condition. The post describes two paths: Path 1 uses Claude Code with Mantle (set CLAUDE_CODE_USE_MANTLE=1) and works in seven Regions (Ireland, Stockholm, Tokyo, Melbourne, US East N. Virginia, US East Ohio, US West Oregon) with native single-Region routing via AWS_REGION, but not in London. Path 2 uses classic Amazon Bedrock (set CLAUDE_CODE_USE_BEDROCK=1) with application inference profiles; it is the only option for London, where Mantle offers no in-Region routing. In London, in-Region support on classic Bedrock is limited to Claude Opus 4.6 and Sonnet 4.6. The IAM policy includes an aws:RequestedRegion condition to restrict calls to the target Region. Verification involves checking CloudTrail events: Path 2 logs InvokeModel under bedrock.amazonaws.com, Path 1 logs CreateInference under bedrock-mantle.amazonaws.com; each event's awsRegion must match the target, and queries in other Regions must return zero. The post notes that for most workloads cross-Region inference is the default, but single-Region is for specific compliance requirements. Cleanup involves deleting application inference profiles via AWS CLI.
Abbreviations
AWS = Amazon Web Services — Amazon Web Services
IAM = Identity and Access Management — Identity and Access Management
CLI = Command Line Interface — Command Line Interface
ARN = Amazon Resource Name — Amazon Resource Name
Source: AWS ML blog — original
Our earlier posts on this topic ↓
Fresh news