---
type: "post"
id: "build-a-semantic-layer-for-agentic-ai-on-b577b3"
slug: "build-a-semantic-layer-for-agentic-ai-on-b577b3"
title: "Build a semantic layer for agentic AI on AWS with AtScale and Amazon Bedrock AgentCore"
description: "**Thesis:** Agent access to enterprise analytics is more governable when the agent queries a shared semantic model whose definitions and permissions are enforced outside the prompt."
retrieval_nugget: "**Thesis:** Agent access to enterprise analytics is more governable when the agent queries a shared semantic model whose definitions and permissions are enforced outside the prompt."
published_at: "2026-10-20"
updated_at: "2026-10-20"
record_date: "2026-10-20"
date_kind: "published_at"
topics: ["mcp"]
source_urls: ["https://aws.amazon.com/blogs/machine-learning/build-a-semantic-layer-for-agentic-ai-on-aws-with-atscale-and-amazon-bedrock-agentcore"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:build-a-semantic-layer-for-agentic-ai-on-b577b3"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:1bb9cb97b9519e8ebda9e0c06ff3ec6e1e6623bbf6ce8f23fb915cd287ab0c9d","reviewed_at":"2026-10-20T20:08:49.240Z","source_evidence_count":1,"verified_claim_count":3,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":3,"implication_count":2,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"Agent access to enterprise analytics is more governable when the agent queries a shared semantic model whose definitions and permissions are enforced outside the prompt.","observed_facts":[{"text":"AWS says one deployed AtScale dimensional model is exposed through SQL, XMLA, and MCP for both an agent and BI tools.","source_urls":["https://aws.amazon.com/blogs/machine-learning/build-a-semantic-layer-for-agentic-ai-on-aws-with-atscale-and-amazon-bedrock-agentcore"]},{"text":"The walkthrough places joins, metric definitions, and row and column filters in the semantic layer rather than the prompt.","source_urls":["https://aws.amazon.com/blogs/machine-learning/build-a-semantic-layer-for-agentic-ai-on-aws-with-atscale-and-amazon-bedrock-agentcore"]},{"text":"AWS describes human identity propagation through AgentCore Identity to AtScale, where RBAC and row security are enforced at query time.","source_urls":["https://aws.amazon.com/blogs/machine-learning/build-a-semantic-layer-for-agentic-ai-on-aws-with-atscale-and-amazon-bedrock-agentcore"]}],"mechanism":"The semantic model mediates between agent query planning and governed data access: Strands uses MCP while AtScale applies declared metrics, joins, and caller-specific controls; AgentCore hosts the agent and manages its outbound credential.","why_now":"As teams connect agents to business data, they need shared metric definitions and access rules across agent and conventional analytics interfaces.","implications":["Teams can evaluate a governed multi-protocol semantic model as a way to reuse established metric logic and permissions for agent workflows.","Policy enforcement outside the prompt provides a more inspectable data-facing boundary, but does not itself establish correctness or security."],"evidence_boundary":"The supplied AWS source is a technical walkthrough, not an independent security assessment, benchmark, production outcome, or comparative evaluation.","watch_conditions":["Look for production audits of agent queries and tests that row-level access remains correct across user roles."],"related_records":[],"new_branch_reason":"No prior New Runtime analytical record is supplied for this specific AtScale implementation; it merits a distinct record about semantic-model mediation and query-time identity controls."}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/build-a-semantic-layer-for-agentic-ai-on-b577b3/","markdown":"https://newruntime.com/posts/build-a-semantic-layer-for-agentic-ai-on-b577b3.md","json":"https://newruntime.com/posts/build-a-semantic-layer-for-agentic-ai-on-b577b3.json"}
related_shifts: ["shared-context-infrastructure"]
---

# Build a semantic layer for agentic AI on AWS with AtScale and Amazon Bedrock AgentCore

## Retrieval answer

**Thesis:** Agent access to enterprise analytics is more governable when the agent queries a shared semantic model whose definitions and permissions are enforced outside the prompt.

**Thesis:** Agent access to enterprise analytics is more governable when the agent queries a shared semantic model whose definitions and permissions are enforced outside the prompt.

AWS’s AtScale walkthrough grounds that claim in a specific architecture. **Observed fact 1:** one deployed dimensional model is exposed over SQL, XMLA, and MCP, so the agent and existing BI tools use common definitions. **Observed fact 2:** the example places joins, metric definitions, and row and column filters in the semantic layer rather than in the prompt. **Observed fact 3:** AWS describes propagating human identity through Bedrock AgentCore Identity to AtScale, where RBAC and row security are applied at query time.

**Mechanism:** The semantic model mediates between agent query planning and governed data access: Strands uses MCP while AtScale applies declared metrics, joins, and caller-specific controls; AgentCore hosts the agent and manages its outbound credential.

**Why now:** As teams connect agents to business data, they need shared metric definitions and access rules across agent and conventional analytics interfaces.

**Implication:** Teams can evaluate a governed multi-protocol semantic model as a way to reuse established metric logic and permissions for agent workflows. This is an architectural option, not proof that every agent query will be correct or that the setup is simpler in every environment.

**Evidence boundary:** The supplied AWS source is a technical walkthrough, not an independent security assessment, benchmark, production outcome, or comparative evaluation.

**Watch condition:** Look for production audits of agent queries and tests that row-level access remains correct across user roles.

The example’s architecture makes a distinction that matters in deployment: an agent may plan a query, but a separate data-facing layer defines what that query means and which results its caller can see. AgentCore hosts the Strands agent and manages its outbound credential, while AtScale supplies shared business definitions and applies controls when a query runs. The source does not establish how this arrangement performs across workloads or whether it reduces operational effort. The next useful evidence would come from production use: query audits, role-based access tests, and confirmation that agents and BI users continue to receive consistent metric definitions.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Mcp](https://newruntime.com/topics/mcp.md): Explore the mcp topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)