Oleksandr Kuzmenko, EditorPublished 5 min readAI-assisted · editor-reviewed
TL;DR
- Bedrock traces inference calls to IAM principals without requiring custom proxies or logging instrumentation.
- IAM Identity Center SSO and STS session tags pass user identity context directly into runtime model calls.
- Static IAM credentials bypass per-user tracking; migration to short-lived SSO session profiles is required.
Key facts
- Instrumentation
- Zero-instrumentation built-in attribution
- Identity Provider
- AWS IAM Identity Center
- Mechanism
- AWS STS session tags mapped to inference calls
Zero-Instrumentation Cost Attribution
Amazon Bedrock provides built-in cost attribution that traces every model inference call back to the specific AWS IAM principal executing the request. Previously, organizations running multi-tenant AI developer tools or shared accounts had to deploy custom API gateways, reverse proxies, or dedicated logging layers to calculate per-user spend.
Integrating IAM Identity Center and Session Tags
The attribution mechanism works directly with AWS IAM Identity Center. When developers authenticate into local CLI or developer tools via single sign-on, short-lived credentials carry identity context into AWS STS session tags. Bedrock captures these session tags during runtime model invocations, attaching the identity to model usage without modifying application code.
Operational Caveats for AI Workflows
To take advantage of this feature, AI engineering leads must audit how local CLI tools and agentic environments connect to AWS. Tools configured with long-lived IAM user keys or shared role credentials will group all token spend under a single bucket. Switching developer setups to dynamic, short-lived SSO profiles enables visibility into individual engineer spend across AI productivity tools.
When to use
- When tracking and attributing Bedrock LLM consumption across multiple engineers using local AI developer tools.
When NOT to use
- When development workflows rely entirely on static, shared API credentials without IAM Identity Center SSO integration.
What to do today
- Audit local CLI and IDE setups to replace static AWS access keys with SSO profiles backed by IAM Identity Center.
- Configure AWS STS session tags to propagate developer identifiers to Bedrock inference requests.
Step names are taken from What to do today.
Step names are taken from What to do today. Audit local CLI and IDE setups to replace static AWS access keys with SSO profiles backed by IAM Identity Center.. Configure AWS STS session tags to propagate developer identifiers to Bedrock inference requests.
- Audit local CLI and IDE setups to replace static AWS access keys with SSO profiles backed by IAM Identity Center.
- Configure AWS STS session tags to propagate developer identifiers to Bedrock inference requests.
