How Can AgentCore Query Cross-Account DynamoDB While Preserving User Identity?

Implement data security and privacy controls. Implement agentic AI solutions and tool integrations. Design and implement enterprise integration architectures.
Answer Correct answer: A — Use an AgentCore outbound credential provider to AssumeRole into per-account IAM roles with session tags, preserving user identity and audit trails.

A global company uses an organization in AWS Organizations to manage 25 AWS accounts. The company’s products have 50,000 daily user interactions. The company is using Amazon Bedrock AgentCore to build an AI assistant. The AI assistant must access customer data that is stored in multiple AWS accounts across sever-al business units. The AI assistant needs to perform actions on behalf of specific users by query-ing Amazon DynamoDB tables that contain customer information in each business unit's AWS account. Each business unit maintains its own AWS account that hosts its own DynamoDB ta-bles. The company must enforce strict data access security boundaries and must collect full audit trails that show which user’s credentials are used to perform each action. The solution must allow the AgentCore application to access resources across multiple accounts while preserving user identity context. The solution must support the company’s plan to expand to more than 40 AWS accounts in the next year. Which solution will meet these requirements with the LEAST operational overhead?

  1. Configure the AI assistant to use an outbound credential provider that runs AssumeRole op-erations to dynamically obtain temporary credentials so the AI assistant can access resources across accounts. Create an IAM role in each business unit account that has a trust relationship to the AI assistant’s execution role and permissions to access the DynamoDB tables. Ensure that AssumeRole calls include session tags to identify the end user. Correct Answer
  2. Create a separate AI assistant in each business unit account. Deploy an Amazon API Gateway REST API that runs request routing logic. Create AWS Lambda authorizers to direct user queries to the appropriate AI assistant based on the customer’s business unit identifier from the request context.
  3. Use AWS Resource Access Manager (AWS RAM) to share the DynamoDB tables from each business unit account to a central account where the AI assistant runs. Configure AWS RAM permissions to allow cross-account access to the DynamoDB tables. Grant the AgentCore execu-tion role permissions to access the shared resources directly.
  4. Configure SCPs for the organization to explicitly allow the AI assistant’s execution role prin-cipal to perform dynamodb:Getltem and dynamodb:Query actions across all member accounts. Configure the AI assistant to reference cross-account DynamoDB table ARNs directly in their resource definitions.

Community Insight

The question tests cross-account access design for AgentCore with per-user auditability; the trap is thinking SCPs grant permissions or that AWS RAM sharing carries end-user identity context.

This scenario asks how an Amazon Bedrock AgentCore assistant should query DynamoDB tables in many AWS accounts while preserving each end user's identity and audit trail. The correct design uses an outbound credential provider with cross-account AssumeRole calls and IAM session tags, which scales to 40+ accounts with minimal operational overhead.

A frequent wrong pick is C, assuming AWS RAM is the right way to centralize DynamoDB access and that shared-resource permissions satisfy the audit requirement; RAM sharing does not preserve the end user's credential context, and it adds share-management overhead across business units.

Community Discussion (2 comments)

TNT87 👍 1 Selected: A
Cross-Account Access: Implementing AssumeRole allows the central AI assistant to securely access DynamoDB tables residing in multiple distinct member accounts within the AWS Organization without requiring separate agent deployments
TNT87 👍 1 Selected: A
Least Operational Overhead & High Scalability: Centralizing the AI assistant using Amazon Bedrock AgentCore allows the architecture to easily scale from 25 to more than 40 accounts. Instead of deploying and managing duplicate agent configurations across dozens of business unit accounts (as proposed in Option B), a single assistant can orchestrate cross-account workflows

Comments & Corrections

No comments yet — spotted an error or have a note? Share it below.

Log in to comment, report an error, or add a note about this question.

Submitted for moderation before publishing. Keep it helpful and respectful.

Expert Analysis

Why the Answer Is Correct

Option A is correct because AgentCore's outbound credential provider can assume an IAM role in each business unit account, so the assistant uses short-lived credentials scoped to that account's DynamoDB tables. Session tags passed during AssumeRole carry the end-user identity into the assumed-role session, and CloudTrail records the role session and tags for audit trails. This centralizes one assistant and scales by adding trust and permissions in new accounts, which is far less overhead than duplicating agents. TNT87's point that "a single assistant can orchestrate cross-account workflows" matches the requirement to expand beyond 40 accounts.

Why the Other Options Are Wrong

B deploys a separate AI assistant in each business unit plus API Gateway routing and Lambda authorizers, creating duplicate agent configurations and fragmented audit trails. C relies on AWS RAM to share DynamoDB tables centrally, but resource sharing does not provide the per-user credential context needed for full auditability and adds share-management overhead across business units. D uses SCPs, which only set the maximum permissions for accounts in AWS Organizations and never grant access or create cross-account trust. Directly referencing cross-account DynamoDB table ARNs in resource definitions also does not establish the required role trust or user identity context.

Community Comment Notes

The community discussion by TNT87 strongly favors option A, noting that AssumeRole lets the central AI assistant access DynamoDB tables in multiple distinct member accounts without separate agent deployments. TNT87 also emphasizes least operational overhead and high scalability, observing that "a single assistant can orchestrate cross-account workflows." That aligns with the requirement to grow from 25 to more than 40 AWS accounts. No comments argue for B, C, or D.

Official Reference

Exam Strategy

For cross-account AgentCore questions, eliminate options that duplicate the agent, rely on SCPs as grants, or share resources without user identity context. Look for AssumeRole plus session tags when the requirement mentions per-user credentials and audit trails.

Frequently Asked Questions

Why does AWS RAM not satisfy the DynamoDB cross-account requirement?

RAM shares resources, not the end user's credentials; even if a table were shared, DynamoDB access would not carry the per-user session tags needed for a full audit trail.

Why can't SCPs grant AgentCore access to DynamoDB in other accounts?

SCPs only define the maximum permissions for accounts in Organizations; they do not grant permissions or create cross-account trust. AgentCore still needs an assumed IAM role in each account.

More AIP-C01 FAQ →

Related Analysis

A financial services company uses Amazon Bedrock to analyze

A-Grade

A financial services company runs an AI application that use

S-Grade

A company has deployed an AI assistant as a React applicatio

S-Grade

Practice All AIP-C01 Questions

Access 110 questions with complete answers and detailed explanations.

View Full AIP-C01 Practice Test →

← Back to AIP-C01 Study Guide