MoEngage Transforms Over 400 APIs into Agent Tools with AgentCore Gateway, Boosting Platform Efficiency

Summary:MoEngage has integrated Amazon Bedrock's AgentCore Gateway to transform approximately 400 internal REST APIs into permission-scoped agent tools, avoiding public exposure and context bloat. This approach eliminates the need for multiple MCP servers, streamlines API integration, and saves an estimated six months of engineering time. It also significantly accelerates the onboarding of new APIs, reducing deployment time from days to hours. This innovation highlights the potential of AI platforms com

---

Core Breakthrough

MoEngage, through integration with Amazon Bedrock's AgentCore Gateway, successfully transformed approximately 400 internal REST APIs into agent tools. The core breakthroughs of this solution include:

1. Avoiding Public Exposure: All APIs run within private VPCs, eliminating the need for public endpoints or additional authentication mechanisms, ensuring data security.
2. Simplifying Architecture: There is no need to build separate MCP servers or network configurations for each API, significantly reducing operational complexity.
3. Efficient Deployment: The onboarding of new APIs is reduced from days to hours, as the platform team only needs to write OpenAPI specifications and apply Cedar policies.
4. Semantic Search: The AgentCore Gateway supports semantic search, allowing agents to find the right tool by describing the task, without memorizing tool names.

Technical Mechanism Analysis

The AgentCore Gateway achieves these functionalities through the following mechanisms:

• OpenAPI Specification-Driven: The OpenAPI specifications maintained by the team are directly converted into MCP tools, eliminating the need for additional MCP server development.
• Cedar Policy Engine: Before the tool catalog is serialized, the AgentCore policy engine filters tools based on IAM roles, ensuring that agents can only see authorized tools.
• Amazon VPC Lattice Managed Egress: The AgentCore Gateway communicates with internal services via VPC Lattice, avoiding complex network configurations.
• Semantic Search Capability: Agents can find the right tool by describing the task, rather than relying on tool names, enhancing the flexibility and accuracy of tool discovery.

Engineering Trade-offs and Performance

• Advantages:
  • Simplifies the API integration process and reduces operational costs.
  • Improves the efficiency of agent tool discovery and usage.
  • Enhances security by avoiding the public exposure of APIs.
• Challenges:
  • The accuracy of semantic search depends on the quality of tool descriptions, requiring teams to write high-quality descriptive text.
  • The complexity of Cedar policies may increase the difficulty of policy management, requiring platform teams to have relevant expertise.

Developer Implementation and Deployment Recommendations

1. Define API Exposure Scope Clearly: In the OpenAPI specification, clearly define which operations should be exposed as tools and use the internal and external flags to control access scope.
2. Write High-Quality Tool Descriptions: Avoid reusing API descriptions directly; instead, write model-facing descriptive text to improve the accuracy of tool selection.
3. Leverage CI/CD Pipelines for Automated Deployment: Integrate the AgentCore Gateway into the CI/CD pipeline to achieve an automated process from specification to running tool.
4. Choose the Right Authentication Mechanism: Select SigV4 or OIDC JWT for authentication based on the type of agent, ensuring security and ease of use.

Conclusion

MoEngage's AgentCore Gateway solution demonstrates the significant potential of combining AI platforms with cloud services. By separating API ownership and protocol layer management, teams can build and manage agent tools more efficiently while maintaining security and flexibility. This innovation provides a replicable architectural model for other enterprises, helping to accelerate the adoption of AI agents in enterprise applications.

---

Source: AWS Machine Learning Blog (2026-10-29)

— END —

Tags: #MoEngage Follow #AgentCore Gateway Follow #Agent Tools Follow #Amazon Bedrock Follow #AI Platform Follow