AWS Releases Agentic AI Semantic Layer Solution with EDB Postgres AI and Amazon Bedrock AgentCore

Summary:AWS has launched a semantic layer solution for agentic AI built on EDB Postgres AI and Amazon Bedrock AgentCore. This solution embeds a Semantic Knowledge Base (SemKB) within the Postgres database, enabling semantic understanding and access control for business data. Utilizing the AIDB extension, the semantic layer coexists with the data and enforces fine-grained permission management through Postgres' GRANT and row-level security mechanisms. AgentCore serves as the runtime environment, handling

---

Core Architecture and Features

AWS's newly released agentic AI semantic layer solution integrates EDB Postgres AI and Amazon Bedrock AgentCore to provide an efficient and secure data access and governance mechanism for intelligent agents. The key components and functionalities of this solution are as follows:

1. Semantic Knowledge Base (SemKB):

   • Construction: Built within the Postgres database using the AIDB extension. SemKB combines business data with semantic information, defining data meanings and query paths through COMMENT ON statements and aidb.create_semantic_alias commands.
   • Advantages: The semantic information coexists with the data, eliminating the need for additional system deployment and data synchronization.
   • Automated Processing: Supports real-time (Live) and background (Background) processing modes, automatically responding to database schema changes.

2. Access Control and Governance:

   • Postgres Permission Management: Utilizes Postgres' GRANT and row-level security mechanisms to enforce fine-grained permission control for different intelligent agent roles.
   • MCP Protocol: AgentCore communicates with AIDB via the Model Context Protocol (MCP), transmitting authentication information and query requests.
   • Audit Trails: AIDB captures OpenTelemetry trace data, recording SQL executions and tool calls for subsequent auditing and analysis.

3. Agent Runtime (AgentCore):

   • Scheduling and Execution: Handles agent scheduling, query planning, and execution.
   • Tool Invocation: Invokes external tools and services, such as Amazon Bedrock's Claude model, via the MCP protocol.
   • Security: Uses AgentCore Identity to manage tool invocation credentials, ensuring their security.

4. Integration and Scalability:

   • Multi-Data Source Support: SemKB integrates data from different sources through Postgres' foreign keys and views.
   • Analytics Acceleration: For workloads exceeding the operational database's capacity, EDB Postgres Analytics Accelerator (PGAA) provides support for Apache Iceberg, Delta Lake, and Parquet data formats.

Engineering Trade-offs and Performance

• Advantages:

  • Semantic Understanding: SemKB enables intelligent agents to understand the business meaning of data, improving query accuracy and the naturalness of agent interactions.
  • Security: Fine-grained permission control and data access auditing ensure data security and compliance.
  • Scalability: Supports multi-data source and multi-agent scenarios, adapting to complex business needs.

• Challenges and Trade-offs:

  • Complexity: Building and maintaining SemKB requires deep business understanding and meticulous data model management.
  • Performance Overhead: While AIDB and AgentCore are optimized, high-concurrency query scenarios may introduce some performance overhead.
  • Dependency: The solution heavily relies on the Postgres database and the AWS ecosystem, which may pose migration costs for non-AWS users.

Developer Deployment Recommendations

1. Data Preparation: Select the Postgres database that intelligent agents need to access and create a dedicated data schema.
2. Semantic Annotation: Use COMMENT ON statements to add business meaning annotations to tables, views, and columns, and define common query paths as semantic aliases.
3. Permission Configuration: Configure Postgres permissions according to intelligent agent roles, ensuring the principle of least privilege.
4. Deploy AgentCore: Deploy AgentCore in the AWS environment and configure the MCP protocol and OpenTelemetry tracing.
5. Testing and Validation: Validate the effectiveness of semantic understanding and permission control by simulating intelligent agent queries and adjust configurations as needed.
6. Monitoring and Optimization: Use OpenTelemetry trace data to monitor intelligent agent behavior and optimize based on performance metrics.

Conclusion

AWS's integration of EDB Postgres AI and Amazon Bedrock AgentCore provides an innovative solution for the semantic layer of agentic AI. This solution combines semantic understanding, fine-grained permission control, and efficient data access mechanisms, offering strong support for the application of intelligent agents in AI. Despite some challenges in complexity and performance, the solution holds significant application value for enterprise-level applications that require high security and scalability.

---

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

— END —

Tags: #AWS Follow #EDB Postgres AI Follow #Amazon Bedrock AgentCore Follow #Agentic AI Follow #Semantic Layer Follow