JA EN
shiichan

AWS Launches Context Ontology Accelerator to Give AI Agents Real Business Context

AWS What's New ★★★★ View Markdown
#aws#ai#infrastructure

Hi everyone, it's Shii-chan! Today AWS quietly dropped something that matters a lot if you're trying to run AI agents in production — a new open source tool that teaches agents your company's actual rules. Let's dive in!

AWS What's New aws.amazon.com

What was announced?

According to AWS What's New, AWS has released Context Ontology Accelerator, an open source accelerator that helps you build an ontology of your business — a machine-readable model of your products, customers, policies, and the rules that define how you operate. The goal is to help AI agents make decisions that are accurate, consistent, explainable, and auditable.

Why it matters

Teams building AI agents want to trust the decisions those agents make. The problem is that the context agents need — entities, rules, policies, relationships — is scattered across dozens of systems with conflicting schemas and naming conventions.

As a result, teams spend months reconciling definitions and hand-coding them into prompts and one-off integrations. And because the outcome isn't traceable, many teams end up blocking their agents from ever moving out of proof-of-concept into production. Context Ontology Accelerator exists specifically to break that logjam.

What changes

With Context Ontology Accelerator, building an ontology looks like this:

  • Connect it to your structured and/or unstructured data sources
  • AI drafts an initial ontology for you
  • Your domain experts review, edit, and approve every element
  • The approved ontology is stored as a knowledge graph that you own
  • It's expressed in open W3C standards, so it works with any standards-based tooling
  • Any agent can consume it through the included Model Context Protocol (MCP) server

The headline benefit: work that used to take months of manual ontology authoring can now go from connecting your data to serving governed answers to agents in a matter of days.

Dive Deep

Under the hood, the accelerator leans on existing AWS managed services:

  • Graph store: Amazon Neptune
  • Vector store: Amazon OpenSearch Serverless
  • Foundation models: Amazon Bedrock

The accelerator is available today on GitHub under the Apache 2.0 license, so you can self-deploy it, or bring in AWS Professional Services to help set it up.

Looking ahead, this ontology capability is expected to become a native, managed feature inside AWS Context, and ontologies you build with the accelerator today should be able to migrate over.

Wrap-up

  • AWS released Context Ontology Accelerator, an open source tool for building business ontologies
  • Flow: connect data → AI drafts → experts review and approve → stored as a W3C-standard knowledge graph
  • The included MCP server lets any agent query the ontology
  • Under the hood: Amazon Neptune, Amazon OpenSearch Serverless, and Amazon Bedrock
  • Available today on GitHub under Apache 2.0

If you're building enterprise AI agent infrastructure and need decisions you can actually explain and audit, this one's worth a look!

Related posts