AWS Releases Amazon Bedrock AgentCore Gateway with OBO Token Exchange Support for ServiceNow

Summary:AWS has introduced a new feature for Amazon Bedrock AgentCore Gateway that enables On-Behalf-Of (OBO) token exchange with Microsoft Entra ID. This allows AI agents to call the ServiceNow REST API while preserving the identity of the calling user, ensuring user-level access control and maintaining audit trails. The feature is natively supported by AgentCore Gateway, eliminating the need for custom interceptors or code, thereby reducing development and maintenance overhead.

---

Core Features and Architecture

The latest update to Amazon Bedrock AgentCore Gateway introduces native support for OBO (On-Behalf-Of) token exchange, enabling AI agents to securely call protected REST APIs while preserving user identity. The key architectural components include:

1. AI Agent Application: Protected via OAuth application registration with Microsoft Entra ID.
2. AgentCore Gateway (MCP Server): Acts as an intermediary between the AI agent and ServiceNow, handling token exchange.
3. ServiceNow REST API: Configured to accept OAuth integrations from Entra ID, ensuring correct user identity mapping.

Technical Mechanism Analysis

The core mechanism of the OBO token exchange is as follows:

• Initial Token Acquisition: The user obtains a JWT (JSON Web Token) to invoke the AI agent.
• Agent Authentication and Token Exchange: The agent authenticates with the AgentCore Gateway using the JWT, and the Gateway exchanges the token via the OBO flow for a new token scoped to the ServiceNow API, preserving the original user identity.
• ServiceNow Access: ServiceNow receives a token containing the user identity and applies the corresponding permissions.

Engineering Trade-offs and Performance

• Advantages: Eliminates the need for custom Lambda or interceptor code, simplifying development and reducing operational costs.
• Challenges: Requires correct configuration of Entra ID and ServiceNow OAuth integrations to ensure that tokens contain necessary user claims (e.g., preferred_username).
• Performance: In testing, the AgentCore Gateway efficiently handles high-concurrency requests with negligible latency increases.

Developer Onboarding and Deployment Recommendations

1. Configure Microsoft Entra ID: Create application registrations, configure the user_impersonation scope, and ensure the AI agent application has permissions to call the ServiceNow API.
2. Configure ServiceNow: Set up inbound OAuth integrations, configure the OIDC provider, and map user claims to ServiceNow user fields.
3. Configure AgentCore Gateway: Create outbound authentication configurations, select JWT authorization grants, and configure token exchange parameters (e.g., requested_token_use: on_behalf_of).
4. Test and Validate: Before deployment, simulate user requests to verify that token exchange and permission application are functioning correctly.

Conclusion

The OBO support in AgentCore Gateway provides a more secure and efficient way for AI agents to integrate with protected REST APIs. It eliminates the need for custom code and ensures that user identity is preserved throughout the call chain, offering stronger security and scalability for enterprise AI applications.

Tags

• AWS
• Amazon Bedrock
• AgentCore Gateway
• OBO Token Exchange
• AI Agent
• ServiceNow

Reading Time

5 minutes

---

Source: AWS Machine Learning Blog (2026-11-03)

— END —

Tags: #AWS Follow #Amazon Bedrock Follow #AgentCore Gateway Follow #OBO Token Exchange Follow #AI Agent Follow #ServiceNow Follow