AWS Details Two Paths for Single-Region Claude Code Deployments on Amazon Bedrock
AWS published a technical guide detailing two methods for keeping Claude Code inference confined to a single AWS Region: Anthropic's newer Mantle endpoint and the classic Bedrock Invoke API with application inference profiles. The right path depends entirely on which Region compliance teams require.
AWS has published guidance for enforcing strict data-residency requirements when running Claude Code on Amazon Bedrock, addressing a gap that arises when compliance teams demand inference stay within a single named AWS Region rather than a broader geography.
The post stems from a real customer request: a US-headquartered global organization needed Bedrock model inference for Claude Code processed specifically in London (eu-west-2), not merely invoked from London. AWS engineers found that Anthropic's newer Mantle endpoint on Bedrock could not satisfy this requirement for that Region, so they fell back to the classic Bedrock Invoke API combined with an application inference profile and an IAM Region condition.
Two endpoints, two mechanisms
Bedrock exposes Claude through two distinct endpoints, each with different Regional coverage:
-
Classic Bedrock (bedrock-runtime): Activated via
CLAUDE_CODE_USE_BEDROCK=1. Single-Region routing requires creating an application inference profile pointed at an in-Region foundation model, since system-defined profiles default to cross-Region routing. Per AWS, eu-west-2 (London) currently supports only Claude Opus 4.6 and Sonnet 4.6 for in-Region inference on this path, using model IDsanthropic.claude-opus-4-6-v1andanthropic.claude-sonnet-4-6. -
Mantle (bedrock-mantle): Activated via
CLAUDE_CODE_USE_MANTLE=1. This is Anthropic's native API shape served through Bedrock, requiring only thatAWS_REGIONbe set — no inference profile needed. Mantle supports native in-Region routing in seven Regions: Ireland, Stockholm, Tokyo, Melbourne, US East (N. Virginia), US East (Ohio), and US West (Oregon), with access to newer models including Claude Sonnet 5, Opus 4.8, and Haiku 4.5.
Critically, Mantle has no in-Region-only option for eu-west-2 — only Global and EU-wide endpoints, meaning a London request could be processed in Frankfurt, Ireland, or Paris. For customers requiring London specifically, that disqualifies Mantle entirely, leaving classic Bedrock with an application inference profile as the only compliant route.
AWS requires Claude Code v2.1.94 or later for Mantle support. Both paths use IAM Region conditions (aws:RequestedRegion) attached to the developer's IAM role or user to reject any call not targeting the approved Region, even if environment variables are misconfigured. AWS recommends verifying compliance through CloudTrail logs.
AWS also notes an important caveat: for most Bedrock workloads, cross-Region inference (CRIS) remains the recommended default, since it improves throughput, capacity, and access to new models. The single-Region patterns described apply only when compliance mandates a specific Region rather than a general geography — AWS suggests EU cross-Region profiles suffice when "somewhere in the EU" is an acceptable boundary.
What this means
This is an operational fix for enterprise compliance, not a new model or pricing change. It signals that data-residency enforcement on Bedrock is still uneven across Regions and endpoints — Mantle, Anthropic's newer native-API path, currently cannot guarantee in-Region-only processing in London, forcing reliance on the older Invoke API and manually configured inference profiles. Enterprises with strict per-Region requirements (finance, healthcare, government contractors) need to audit which endpoint and Region combination actually satisfies their obligations before deploying Claude Code, since the two paths differ in available models, IAM permissions, and enforcement mechanics. AWS's explicit advice to escalate to account teams for unsupported Regions suggests this remains a manual, case-by-case problem rather than a fully generalized platform capability.
Related Articles
Anthropic Reports Bacterial and Phage Proteins Assemble Into Vault Particles in E. coli, With Claude Aiding a 3.2 Å Map
Anthropic reported on November 4, 2026 that prokaryotic and bacteriophage relatives of the major vault protein can assemble into vault particles when produced in E. coli. Claude helped the team map one bacteriophage-derived shell at a nominal resolution of 3.2 angstroms.
Anthropic releases FlashProve, open-source cryptographic proofs for AI inference, at 110 tokens per second
Anthropic released FlashProve on November 2, 2026, an open-source research prototype that uses cryptographic proofs to show a language-model response came from committed weights without revealing them. The company reported proof generation at 110 tokens per second for Meta's Llama 4 Maverick.
Moonshot AI launches Kimi Hosted Agents, a managed agent runtime with sandboxes at 0.072 yuan per hour
Moonshot AI has launched Kimi Hosted Agents, a Harness as a Service product that runs the agent runtime on Kimi's infrastructure. It offers isolated sandboxes, exportable audit logs, more than 140 skills and more than 30 plugins. IT Home cites a sandbox rate of 0.072 yuan per hour plus model and plugin charges.
Huawei's Xiaoyi Work is first to run openPangu-2.1-Pro, labeled for HarmonyOS task enhancement
Huawei's Xiaoyi Work AI work agent is the first product to offer openPangu-2.1-Pro, a new version of the open Pangu model family, according to IT Home's hands-on testing on November 5. Huawei has not published parameter count, context length or a release date.
Comments
Loading...