---
type: "post"
id: "build-a-semantic-layer-for-agentic-ai-on-ec1cc0"
slug: "build-a-semantic-layer-for-agentic-ai-on-ec1cc0"
title: "Build a semantic layer for agentic AI on AWS with Strategy Mosaic and Amazon Bedrock AgentCore"
description: "The thesis is that Governed semantic definitions can act as the stable boundary between enterprise data and an agent, rather than asking the agent prompt to carry business logic and access rules."
retrieval_nugget: "The thesis is that Governed semantic definitions can act as the stable boundary between enterprise data and an agent, rather than asking the agent prompt to carry business logic and access rules."
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-strategy-mosaic-and-amazon-bedrock-agentcore"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:build-a-semantic-layer-for-agentic-ai-on-ec1cc0"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:ecf395b5efd0332b40768a1292af4a955034c4297d5ed8845bbb1b380798489c","reviewed_at":"2026-10-20T20:08:49.240Z","source_evidence_count":1,"verified_claim_count":2,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":2,"implication_count":1,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"Governed semantic definitions can act as the stable boundary between enterprise data and an agent, rather than asking the agent prompt to carry business logic and access rules.","observed_facts":[{"text":"The example connects a customer-360 model spanning Amazon Redshift and Amazon Aurora to a Strands agent hosted on Amazon Bedrock AgentCore through Strategy Mosaic’s MCP server.","source_urls":["https://aws.amazon.com/blogs/machine-learning/build-a-semantic-layer-for-agentic-ai-on-aws-with-strategy-mosaic-and-amazon-bedrock-agentcore"]},{"text":"The post says joins, metric definitions, and per-caller row and column filters live in Mosaic, with user identity propagating so those controls apply end to end.","source_urls":["https://aws.amazon.com/blogs/machine-learning/build-a-semantic-layer-for-agentic-ai-on-aws-with-strategy-mosaic-and-amazon-bedrock-agentcore"]}],"mechanism":"The semantic layer owns business definitions and caller-specific constraints; MCP provides the agent access path, while Mosaic and identity propagation are the stated control point.","why_now":"Teams are connecting agents to governed enterprise data and need boundaries that can serve both agent and conventional clients.","implications":["Teams should inspect where metric semantics and authorization are enforced and whether identity survives each hop."],"evidence_boundary":"This is a vendor-authored implementation walkthrough; it does not independently verify security, performance, or generality beyond the demonstrated setup.","watch_conditions":["Whether deployments publish reproducible tests showing caller identity, row filters, and metric definitions remain correct across agent requests and non-agent clients."],"related_records":[],"new_branch_reason":"No prior New Runtime record was supplied; this specific semantic-layer and MCP implementation can be assessed as a distinct record."}
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-ec1cc0/","markdown":"https://newruntime.com/posts/build-a-semantic-layer-for-agentic-ai-on-ec1cc0.md","json":"https://newruntime.com/posts/build-a-semantic-layer-for-agentic-ai-on-ec1cc0.json"}
related_shifts: ["shared-context-infrastructure"]
---

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

## Retrieval answer

The thesis is that Governed semantic definitions can act as the stable boundary between enterprise data and an agent, rather than asking the agent prompt to carry business logic and access rules.

The thesis is that Governed semantic definitions can act as the stable boundary between enterprise data and an agent, rather than asking the agent prompt to carry business logic and access rules. The AWS and Strategy walkthrough offers a concrete instance, not a general proof of the approach.

Observed facts: the example connects a customer-360 model spanning Amazon Redshift and Amazon Aurora to a Strands agent hosted on Amazon Bedrock AgentCore through Strategy Mosaic’s MCP server. Observed facts: the post says joins, metric definitions, and per-caller row and column filters live in Mosaic, and that user identity propagates so those controls apply end to end. It also lists MCP, JDBC/ODBC, XMLA, and REST as interfaces served from the same definition.

The mechanism is that The semantic layer owns business definitions and caller-specific constraints; MCP provides the agent access path, while Mosaic and identity propagation are the stated control point. This is not a claim that MCP itself supplies authorization. This matters now because Teams are connecting agents to governed enterprise data and need boundaries that can serve both agent and conventional clients.

The first implication is that Teams should inspect where metric semantics and authorization are enforced and whether identity survives each hop. The evidence boundary is that This is a vendor-authored implementation walkthrough; it does not independently verify security, performance, or generality beyond the demonstrated setup. The first watch condition is Whether deployments publish reproducible tests showing caller identity, row filters, and metric definitions remain correct across agent requests and non-agent clients.

The story starts a distinct analytical record because no prior New Runtime record was supplied; the source describes a specific semantic-layer/MCP implementation that can be assessed on its own.

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)