Opus 4.8, Sonnet 5, and User Activity Monitoring now available on Kiro in AWS GovCloud (US)
Government teams can now run Opus 4.8 and Sonnet 5 inside GovCloud with full audit-ready usage monitoring and data residency guarantees.
View original announcement →Visual Summary
graph TD
A{{Kiro in AWS GovCloud}}:::announced
B(Amazon Bedrock):::compute
C([Claude Opus 4.8]):::feature
D([Claude Sonnet 5]):::feature
E([User Activity Monitoring]):::feature
F(Amazon S3):::storage
G((Enterprise Admins)):::external
H((Developers)):::external
I([Prompt Logging]):::feature
H ==>|"IDE/CLI access"| A
A -->|"cross-region inference"| B
B -->|"2.2x credits"| C
B -->|"1.3x credits"| D
G -->|"monitors usage"| E
E -->|"daily CSV telemetry"| F
A -.->|"optional"| I
I -.->|"stores logs"| F
G -->|"configures"| A
classDef announced fill:#ff9900,stroke:#ec7211,color:#fff,font-weight:bold
classDef compute fill:#e3f2fd,stroke:#1565c0,color:#1565c0
classDef storage fill:#e8f5e9,stroke:#2e7d32,color:#2e7d32
classDef feature fill:#fff3e0,stroke:#e65100,color:#e65100
classDef external fill:#f5f5f5,stroke:#616161,color:#616161
What's New
AWS has expanded Kiro, its AI-powered development platform, to AWS GovCloud (US) Regions with two new Claude models: Claude Opus 4.8 and Claude Sonnet 5. Both models offer a 1M context window and are accessible via the Kiro IDE and CLI. Alongside the model additions, enterprise administrators gain built-in user activity monitoring capabilities, including a usage dashboard, daily CSV telemetry delivered to S3, and optional prompt logging for compliance and audit purposes.
How It Works
- Claude Opus 4.8 operates with a 2.2x credit multiplier and is designed for complex, long-horizon tasks; it plans before editing, performs self-verification to catch its own errors, and navigates obstacles autonomously rather than stalling.
- Claude Sonnet 5 operates with a 1.3x credit multiplier under an experimental support tag; it delivers near-Opus-4.8 reasoning and agentic coding performance at lower cost, giving teams a tunable cost-performance tradeoff.
- Both models leverage Amazon Bedrock's cross-region inference within GovCloud boundaries: users in AWS GovCloud (US-East) have inference requests processed in AWS GovCloud (US-West), with all cross-region traffic encrypted via TLS 1.2 or higher and data remaining stored in the originating region.
- The usage dashboard provides aggregate organizational metrics at a glance directly within the Kiro Console, enabling administrators to monitor license consumption in real time.
- Per-user activity reports are generated daily as CSV files and delivered to an administrator-specified S3 bucket, containing fields such as credits consumed, model usage, and other telemetry for license optimization and audit trails.
- Prompt logging is an optional feature that captures the full text of user prompts and Kiro responses, storing them in the customer's own AWS account for compliance review and debugging—at no additional charge beyond standard S3 storage costs.
- AWS CloudTrail and Amazon CloudWatch integrations complement native monitoring, enabling API-level audit logging and real-time metric alarms (e.g., daily active users, invocation counts).
Why It's Important
- GovCloud parity for regulated workloads: Federal agencies, defense contractors, and other regulated entities can now leverage state-of-the-art agentic coding models within the compliance boundary of AWS GovCloud (US), where data sovereignty and export-control requirements are enforced.
- Reduced supervision overhead: Opus 4.8's self-verification and autonomous obstacle-navigation capabilities mean complex multi-step development tasks can proceed with less human intervention, directly reducing engineering toil on long-horizon projects.
- Cost flexibility for government teams: The Sonnet 5 / Opus 4.8 pairing gives procurement-conscious government teams a dial between maximum accuracy (Opus 4.8 at 2.2x credits) and higher throughput (Sonnet 5 at 1.3x credits), optimizing budget utilization.
- Audit-ready compliance posture: Built-in prompt logging and daily S3 telemetry make it straightforward to satisfy FedRAMP, FISMA, and internal audit requirements without building custom logging infrastructure.
- Data residency assurance: All monitoring data—usage reports, prompt logs—is stored in the customer's own AWS account and GovCloud region, ensuring no sensitive government data leaves the compliance boundary.
How It's Different
- GovCloud-specific constraints vs. commercial Kiro: Several commercial Kiro features are intentionally disabled in GovCloud—including VS Code plugin support, inline suggestions, Autonomous Agent mode, social/Builder ID login, and automated model selection—to maintain the stricter security posture required by the environment.
- Self-correcting agentic loop (Opus 4.8): Unlike earlier Opus models, Opus 4.8 explicitly plans before making edits and performs iterative self-verification, reducing the need for human review cycles on complex tasks.
- Sonnet 5 cost-performance positioning: Sonnet 5 approaches Opus 4.8 benchmark performance at a 1.3x vs. 2.2x credit multiplier, a differentiated tradeoff not available with prior Sonnet generations.
- Customer-owned telemetry storage: Unlike SaaS monitoring tools that aggregate data in vendor infrastructure, all Kiro monitoring data lands in the customer's own S3 bucket, preserving data ownership and enabling integration with existing SIEM or data lake pipelines.
- No additional monitoring charge: The monitoring and tracking feature set (dashboard, CSV reports, prompt logging) incurs no platform fee beyond standard S3 storage, contrasting with third-party observability tools that charge per seat or per event.
When to Prefer It
- Use Opus 4.8 when tackling large, multi-file refactors, architectural migrations, or feature implementations that span many interdependent components and historically required close developer supervision at each step.
- Use Sonnet 5 when running high-volume agentic coding workflows—such as CI-integrated code review, automated test generation, or batch documentation updates—where throughput and cost efficiency matter more than maximum accuracy.
- Enable prompt logging when your organization must demonstrate compliance with audit requirements (e.g., FedRAMP High, FISMA Moderate/High) that mandate records of AI-generated content and user interactions.
- Configure daily S3 activity reports when you need to optimize seat license allocation across large engineering teams, identify underutilized subscriptions, or produce usage evidence for budget justification cycles.
- Deploy Kiro in GovCloud whenever your project involves Controlled Unclassified Information (CUI), ITAR-regulated data, or any workload that must remain within the AWS GovCloud (US) compliance boundary.
- Use the CloudWatch integration when you want to set automated alerts—for example, triggering a notification if daily active users drop unexpectedly or if credit consumption spikes beyond a budget threshold.
Availability
- General Availability: Claude Opus 4.8 is generally available in Kiro for AWS GovCloud (US) Regions as of July 24, 2026.
- Experimental / Preview: Claude Sonnet 5 is available with an experimental support tag in GovCloud, indicating it may have additional limitations or change before full GA.
- Supported Regions: AWS GovCloud (US-East) (
us-gov-east-1) and AWS GovCloud (US-West) (us-gov-west-1); cross-region inference routes GovCloud (US-East) requests through GovCloud (US-West) for Bedrock inference. - Pricing: Credits-based model; Opus 4.8 carries a 2.2x credit multiplier and Sonnet 5 carries a 1.3x credit multiplier relative to the base Kiro credit unit. Monitoring features have no additional platform charge beyond S3 storage costs.
- Access requirement: Users must update their Kiro IDE or CLI to the latest version and restart to access the new models via the model selector.
- Known limitations in GovCloud: VS Code plugin, inline suggestions, Autonomous Agent, social/Builder ID login, automated model selection ("Auto"), and S3-based user activity reporting (per the GovCloud User Guide) carry specific restrictions—administrators should review the GovCloud documentation for the current feature availability matrix before planning deployments.
