What Is AWS Context? AWS's Knowledge Graph for AI Agents

---

AWS Context is the knowledge graph service AWS announced on June 17, 2026: it maps the relationships across your existing data into a graph and gives your AI agents agentic search over it. AWS names Glue Data Catalog, SageMaker Unified Studio, and Lake Formation as integrations. As of 2026-11-02, AWS lists it as coming soon.

---

Read every row below as AWS’s description of a service you cannot run yet. Quotes come from the AWS Artificial Intelligence Blog unless a row says otherwise.

Attribute	Detail
What it is	AWS service that “maps the relationships across your existing data into a knowledge graph and provides agentic search”
Announced	June 17, 2026, AWS Summit New York (AWS AI Blog, AWS News Blog)
Status	Coming soon, per the AWS AI Blog and the AWS News Blog, both June 17, 2026; rechecked 2026-11-02
Built on	The knowledge graph technology behind Amazon Quick, extended from personal to organizational
AWS integrations named	AWS Glue Data Catalog, Amazon SageMaker Unified Studio, AWS Lake Formation
Agent access	Agentic search APIs and MCP tools, for agents on Amazon Bedrock AgentCore, Amazon EKS, or MCP-compatible frameworks
Permissions	Each call is “designed to inherit the calling user’s IAM and Lake Formation permissions”
Storage format	AWS says it will publish key metadata in Apache Iceberg format in Amazon S3 Tables
Beyond AWS	“designed to connect to third-party catalogs”
Learning	In AWS’s words, it “gets smarter the more your agents use it”
Related release	Context Ontology Accelerator, open source, July 31, 2026

---

What is AWS Context?

AWS Context is a new AWS service that builds an organizational knowledge graph over the data you already have and lets your agents search it at runtime. According to the AWS Artificial Intelligence Blog (2026), it is “a new service that automatically maps the relationships across your existing data into a knowledge graph and provides agentic search.” Agents use it to “access governed data relationships, business rules, and domain knowledge at runtime.”

AWS states the problem in one line: “Agents are only as intelligent as the context they can reason over.”

The graph technology is not new to AWS. AWS Context “extends the same knowledge graph technology that powers Amazon Quick,” per the same post, “extending what was a personal knowledge graph into an organizational one.” Quick’s graph today is “kept for you, per user,” according to the Amazon Quick User Guide. AWS says that graph already serves “hundreds of thousands of users” daily and “processes millions of requests per day.” Those are Quick’s numbers. AWS Context has no published usage yet.

A knowledge graph for AI agents is one way to give agents the context they need. A context graph is the broader idea. The short video below covers the basics.

A graph your agents query is only as trustworthy as the definitions loaded into it. Where those definitions get authored and certified deserves as much planning as the graph.

---

How does AWS Context work?

AWS describes AWS Context in four moves: it infers relationships across your data, stewards review and promote them, agents query the graph through APIs and MCP, and usage feeds back into ranking. Each step below uses AWS’s own description from the AWS Artificial Intelligence Blog.

1. Infer. AWS Context “automatically maps the relationships across your existing data,” and AWS Glue Data Catalog, Amazon SageMaker Unified Studio, and AWS Lake Formation “integrate with the knowledge graph.”
2. Review. Data stewards curate the result by “reviewing inferred relationships, promoting them to production, and attaching domain-specific knowledge like business definitions and usage rules.”
3. Query. Agents call agentic search APIs and Model Context Protocol tools, whether they are built on Amazon Bedrock AgentCore, deployed on Amazon EKS, or running on MCP-compatible frameworks.
4. Rank. Usage feeds back in. AWS describes a graph that updates from use: it “ranks sources by actual usage and shares what it learns across your organization.”

AWS says setup will take “just a few clicks in the AWS Management Console,” and that “There is no infrastructure to provision or retrieval pipeline to build.”

What AWS has described

Dimension	What AWS says
What agents get	“access governed data relationships, business rules, and domain knowledge at runtime”
Human curation	Stewards manage the graph in a console, “reviewing inferred relationships, promoting them to production, and attaching domain-specific knowledge like business definitions and usage rules”
Learning from use	“It ranks sources by actual usage and shares what it learns across your organization”; “when one agent discovers a correct join path or resolves a schema ambiguity, other agents pick it up”
Query engines	“query your context with Amazon Athena, Amazon Redshift, Apache Spark, or any Iceberg-compatible engine”
Audit	“Because access runs through identity, every interaction is auditable.”
Quick users	“Existing Amazon Quick users benefit immediately.”
Ontology roadmap	The Context Ontology Accelerator’s ontology capability “will become a fully managed feature native to AWS Context” (AWS What’s New, July 31, 2026)

Source for every row except the ontology roadmap: AWS Artificial Intelligence Blog, June 17, 2026.

The Iceberg detail deserves a second look. AWS says AWS Context will publish “all key metadata from structured and unstructured sources” to Apache Iceberg in S3 Tables, queryable with Athena, Redshift, Spark or any Iceberg engine. Whether inferred relationships, promoted rules and usage rankings land there too is not stated on the AWS AI Blog as of 2026-11-02; question 8 below asks. The Glue metadata export to S3 Tables, in preview since 2026-08-05, is the part of this pattern you can test today.

Open questions in AWS’s own description

AWS’s launch post describes two curation paths and does not say how they meet. Stewards promote inferred relationships “to production.” Learned fixes, such as a join path, spread to other agents “without requiring a human re-curate the graph.” Which path governs a shared join path is not stated on the AWS AI Blog as of 2026-11-02.

• If a widely used join path is wrong, what stops other agents inheriting it, and how is a shared learning rolled back?
• AWS names “structured and unstructured sources.” Does that cover streaming sources?

The steward step carries the most weight for context engineering on AWS. AWS says stewards review inferred relationships and promote them to production, so what agents receive depends on who reviews, and against which definitions.

---

Is AWS Context available yet? Status and timeline as of November 2026

As of 2026-11-02, AWS lists AWS Context as coming soon on its AI Blog launch post and its Summit roundup. Every status below was rechecked that day against AWS What’s New and the AWS Artificial Intelligence Blog. For preview access, Regions and pricing, ask your AWS account team. Glue’s preview has since expanded (metadata export to S3 Tables, August 5, 2026); AWS Context’s published status has stayed the same.

AWS Context release timeline, from AWS’s own pages

Date	What AWS published	Status words on the page	Source
2026-06-04	SageMaker Data Agent integrates SageMaker Catalog business context	Available “in all AWS Regions where Amazon SageMaker Unified Studio is available”	AWS What’s New
2026-06-16	Amazon S3 annotations, up to 1GB per object	“available in all AWS Regions, including the AWS China Regions”	AWS What’s New
2026-06-17	AWS Context introduced in the AWS Summit New York keynote	“AWS Context (Coming soon)”	AWS AI Blog
2026-06-17	Summit roundup	“Coming soon: AWS Context”	AWS News Blog
2026-06-17	Glue Data Catalog business context and semantic search	“available in preview” in N. Virginia, Ohio, Oregon, and Ireland	AWS What’s New
2026-06-17	Amazon Bedrock Managed Knowledge Base	“now generally available”	AWS What’s New
2026-07-31	Context Ontology Accelerator, open source (Apache 2.0)	“available today on GitHub”; ontology capability “will become a fully managed feature native to AWS Context”	AWS What’s New
2026-08-05	Glue metadata export to S3 Tables; semantic search for catalogs encrypted with customer-managed KMS keys	“(Preview)”, same four Regions	AWS What’s New

The Context Ontology Accelerator is the one development since June that points at AWS Context directly. It is open source on GitHub under the Apache License 2.0, and AWS says its ontology capability “will become a fully managed feature native to AWS Context.” One caution: the What’s New URL for the Accelerator contains the words “generally-available,” but the post describes the open-source release. AWS Context’s own status is unchanged.

Good to know about AWS Context today

• Access runs through identity. AWS says each call is designed to inherit the caller’s IAM and Lake Formation permissions, and that “every interaction is auditable.”
• “Existing Amazon Quick users benefit immediately,” per AWS. When AWS Context is enabled, Quick’s agents “gain access to the broader enterprise knowledge graph.”

For the AWS-native governance services you run today, plan against what has shipped: Glue business context in preview, S3 annotations available in all Regions. Keep AWS Context as an undated roadmap item until AWS publishes Regions and a launch date.

---

How does AWS Context relate to Glue Data Catalog, SageMaker Catalog, Bedrock Knowledge Bases, and Neptune?

AWS Context sits next to several AWS services your team may already run, and AWS names three of them as integrations: Glue Data Catalog, SageMaker Unified Studio, and Lake Formation. Others, such as Amazon Bedrock Knowledge Bases and the Amazon Neptune graph database, solve adjacent problems with their own documentation. The map below keeps each service in AWS’s own words.

The AWS services map

Service	What it is (AWS wording)	Status as of 2026-11-02	How it relates to AWS Context	Source
Amazon Quick knowledge graph	“builds a personal knowledge graph that captures entities and the relationships between them”; “kept for you, per user”	Quick desktop app GA 2026-09-09; the knowledge graph guide page has no status label	AWS Context extends this technology “into an organizational one”	Quick User Guide, AWS What’s New
AWS Glue Data Catalog business context and semantic search	Enrich assets with glossary terms, custom metadata fields, and skill assets; a new Glue Search API	Preview, 4 Regions, “subject to change”	A named integration of AWS Context	Glue Developer Guide
Amazon SageMaker Catalog	“catalog data across your organization with business context”; built on Amazon DataZone	Available; the product page offers “Get started with Amazon SageMaker Catalog”	SageMaker Unified Studio is a named integration of AWS Context	SageMaker Unified Studio guide, product page
AWS Lake Formation	“helps you centrally govern, secure, and globally share data for analytics and machine learning”	Documented in the current Developer Guide; the guide has no status label	A named integration; AWS Context calls are designed to inherit Lake Formation permissions	Lake Formation guide
Amazon Bedrock Knowledge Bases	Managed RAG over your data sources, including documents and structured data stores; Managed Knowledge Base offers “agentic retrieval for multi-hop reasoning” and “integrates natively with AgentCore Gateway so any MCP-compatible agent framework can discover and invoke your Knowledge Base as a tool”	Managed Knowledge Base GA 2026-06-17	Retrieval over your data for agents; AWS describes AWS Context as agentic search over a knowledge graph. Ask AWS how the two divide structured data	Bedrock User Guide, AWS What’s New
Bedrock Knowledge Bases GraphRAG with Neptune Analytics	“a fully managed GraphRAG feature with Amazon Neptune”; “GraphRAG only supports Amazon S3 as the data source”	GA; Regions per the GraphRAG guide	A graph built from documents for RAG; AWS Context maps relationships across data	Bedrock GraphRAG guide, AWS What’s New
Amazon Neptune	“Serverless graph database service for connected data and improved AI accuracy”	Available; the product page offers “Get started with Neptune”	The Ontology Accelerator’s initial release “uses Amazon Neptune as the graph store”	Neptune product page, AWS What’s New
Context Ontology Accelerator	“an open source accelerator that helps organizations build an ontology of their business”; consumable “through the included Model Context Protocol (MCP) server”	Open source, Apache 2.0, on GitHub	Ontology capability slated to become “a fully managed feature native to AWS Context”	AWS What’s New, GitHub
Amazon S3 annotations	Business context on S3 objects, “up to 1GB per object”	Available in all Regions, including China	Same launch wave; S3 Tables is where AWS says AWS Context will publish Iceberg metadata	AWS What’s New
Amazon Bedrock AgentCore	“an agentic platform to build, deploy and operate” agents	GA 2025-10-13	One of the places agents calling AWS Context can run	AWS What’s New

Two preview limits in the Glue guide matter here. During preview, AWS writes, “There is no asset-level access control. Access is managed through IAM,” and “This feature is subject to change,” per the Glue Developer Guide.

Each service answers a different need. Bedrock Knowledge Bases handles managed retrieval over documents and structured stores, and the trade-offs of a Bedrock Managed Knowledge Base or of Bedrock Knowledge Bases vs an external catalog apply today. Neptune stores graphs; comparisons of Amazon Neptune vs a data catalog and of Neptune vs a knowledge graph context layer cover where it fits. On the catalog side, AWS DataZone vs Glue Data Catalog and Amazon DataZone as a data catalog cover Glue Data Catalog, one of AWS Context’s named integrations, and DataZone, which SageMaker Catalog is built on. The Ontology Accelerator adds a modeling step, and the ontology vs knowledge graph distinction explains why that step comes first.

---

What can you use on AWS while AWS Context is coming soon?

Six AWS options can give agents business context today. Their statuses differ, so each row carries one, with its source.

Available today, by need

Need	Available now on AWS	Status	Source
Business definitions on Glue tables for agents	Glue business context, Glue Search API, skill assets; agents connect through the aws-data-analytics plugin	Preview, 4 Regions	AWS What’s New, Glue docs
Context on S3 objects	S3 annotations, queryable through S3 Metadata Iceberg tables and the S3 Tables MCP server	Available	AWS What’s New
An ontology and knowledge graph you own	Context Ontology Accelerator (Neptune, OpenSearch Serverless, Bedrock; MCP server)	Open source; deploy yourself or with AWS Professional Services	AWS What’s New
Retrieval over documents and structured stores; graph RAG	Bedrock Managed Knowledge Base; GraphRAG on Neptune Analytics	GA	Bedrock docs, Managed KB GA, GraphRAG docs, GraphRAG GA
Catalog business context for SageMaker users	SageMaker Catalog; SageMaker Data Agent reads it	Data Agent integration available in all SageMaker Unified Studio Regions	SageMaker docs, AWS What’s New
Context from outside AWS synced into SageMaker Catalog	Third-party catalog connectors that sync into SageMaker Catalog	Data Agent reads synced context in all SageMaker Unified Studio Regions	AWS What’s New

The AWS AI Blog describes the new Glue Search API as a way to “more quickly find data by business meaning,” and skill assets as “a new asset type that references URIs to files.” According to AWS What’s New (2026), any MCP-compatible agent can start with the aws-data-analytics plugin, with virtually no setup.

For teams that want to build a knowledge graph for AI agents now, the Ontology Accelerator is AWS’s own starting point. AWS says it “reduces what would take months of manual ontology authoring into days,” per AWS What’s New (2026). Background on ontology in AI helps scope that work.

SageMaker Data Agent reads business context in SageMaker Catalog, including context synced in from third-party catalogs, per AWS What’s New (2026). An MCP-connected data catalog gives agents a second route to the same definitions.

AWS has committed to one carry-forward path: teams that build ontologies with the Accelerator “will be able to use and manage them with AWS Context.” For Glue business context, AWS names the integration but not what migrates, and the feature is in preview and “subject to change.”

---

How to evaluate AWS Context for your agents: 10 questions to ask AWS

You cannot test AWS Context yet, so the useful evaluation today is a set of questions for your AWS account team, each tied to a specific AWS claim. Every quote below comes from the AWS Artificial Intelligence Blog unless marked.

The 10 questions

1. Is it available to us, and what will it cost? AWS lists it as “Coming soon.” Ask for preview access, target Regions, a date, and how it will be priced, for example whether graph size, query volume or S3 Tables storage is billed. Hold production plans until a What’s New post lists Regions.
2. Which third-party catalogs, and how deep? AWS says it is “designed to connect to third-party catalogs.” Ask which ones, whether glossary terms, owners, and certification state travel, and in which direction.
3. How is learned behavior reviewed? AWS says it “ranks sources by actual usage” and shares join paths “without requiring a human re-curate the graph.” Ask what audit trail exists for an auto-shared join path, and whether stewards can veto it.
4. What do stewards promote, and where? Stewards review “inferred relationships, promoting them to production.” Ask whether promotion happens per relationship or per domain, and whether it is versioned.
5. How do agents outside AWS call it? Access is through “agentic search APIs and MCP tools.” Ask for the MCP tool list and the auth model for agents not running on AgentCore or EKS.
6. How do permissions behave for service identities? Each call is “designed to inherit the calling user’s IAM and Lake Formation permissions.” Ask how an agent that acts for many users gets scoped.
7. How is the graph scoped? Ask whether it is scoped per account, per Region or per SageMaker domain, and how cross-account Lake Formation grants and LF-Tag policies apply to an agent’s call. Lake Formation already lets you share data “across multiple AWS accounts,” per the Lake Formation guide.
8. What lands in Iceberg, and where does it live? AWS says it will publish “all key metadata from structured and unstructured sources.” Ask for the table schema, whether curated rules and usage rankings are included, where the graph and its S3 Tables output reside, and whether customer-managed KMS keys can encrypt them. Glue semantic search added catalogs encrypted with customer-managed KMS keys in preview on 2026-08-05.
9. What carries over from work we do now? Ontology Accelerator users “will be able to use and manage them with AWS Context,” per AWS What’s New, and Glue glossary terms and skill assets sit in a named integration. Ask what migrates automatically.
10. Which glossary is authoritative? Glue now holds glossary terms in preview, and SageMaker Catalog holds business context. Ask which one is authoritative, which one AWS Context reads, and where definitions from outside AWS get written first.

Planning mistakes to avoid while AWS Context is coming soon

• Planning production before Regions exist. Until a What’s New post lists Regions, keep AWS Context as an undated roadmap item.
• Reading status from anywhere but AWS. Check AWS What’s New and the AWS Artificial Intelligence Blog, and date every check.
• Defining business terms three times. Separate definitions in Glue, SageMaker Catalog, and AWS Context give agents three answers. Agree on a source of record before a third copy appears.

A comparison of agent context layer tools frames the wider choice, and a multi-cloud context layer is the pattern for teams whose agents reach past AWS.

---

Where does Atlan fit next to AWS Context?

Atlan, the Context Layer for AI, defines and certifies the context agents read, on AWS and beyond. An enterprise context layer holds those definitions in one Enterprise Data Graph that agents and people share. AWS names Atlan among the third-party catalogs whose context SageMaker Data Agent reads, per AWS What’s New.

What Atlan documents on AWS today:

• AWS connectors. Atlan’s AWS partners page lists data discovery “across Redshift, S3, Glue, and more.” Its 80+ connectors pull context across a data estate.
• SageMaker Catalog sync. AWS documents “On-demand and scheduled bidirectional metadata synchronization” and “Real-time reverse sync of metadata updates from Atlan back to Amazon SageMaker Catalog” on its Atlan integration page. The Atlan connector catalogs domains, projects, business terms, published and subscribed assets, and data products.
• Hosted MCP. Atlan’s MCP server is “hosted and enabled for all Atlan tenants, with nothing to install or run,” and “each person signs in with their own Atlan account,” per Atlan’s docs.

For a primer on the pattern, context layer 101 and Glue and DataZone vs a cross-cloud context layer go deeper.

Once AWS Context ships, your team will still have to answer where a business term is written first, and which copy every other system follows.

---

FAQs about AWS Context

1. Is AWS Context generally available?

No. As of 2026-11-02, AWS lists AWS Context as coming soon, the status it carried in the AWS Artificial Intelligence Blog and the AWS News Blog on June 17, 2026. For preview access, Regions, and pricing, ask your AWS account team.

2. Is AWS Context the same as the Lambda context object?

No. AWS Context is the knowledge graph service for AI agents that AWS announced on June 17, 2026. The Lambda context object is something else: when Lambda runs a function, it “passes a context object to the handler” with information “about the invocation, function, and execution environment,” per the AWS Lambda Developer Guide.

3. How is AWS Context different from Amazon Bedrock Knowledge Bases?

AWS describes Amazon Bedrock Knowledge Bases as managed RAG that can also connect to structured data stores, and its Managed Knowledge Base as offering “agentic retrieval for multi-hop reasoning.” It describes AWS Context as “agentic search” over a knowledge graph of the relationships across your data. Whether one will back or replace the other for structured data is a question to put to AWS.

4. Does AWS Context learn from agent usage?

AWS says it “gets smarter the more your agents use it”: it ranks sources by actual usage and shares fixes, such as a correct join path, across agents. Stewards separately review inferred relationships and promote them to production.

5. Can AWS Context use context from outside AWS?

AWS says AWS Context is “designed to connect to third-party catalogs, so you can bring context from systems beyond AWS into the same graph.” Ask your AWS account team which catalogs, and in which direction context flows.

6. What is the Context Ontology Accelerator?

It is an open-source accelerator AWS released on July 31, 2026 under Apache 2.0 for building a business ontology stored in a knowledge graph you own, which any agent can consume through an included MCP server. AWS says its ontology capability “will become a fully managed feature native to AWS Context.”

---

Sources

1. Context intelligence for your data and AI agents at scale, AWS Artificial Intelligence Blog
2. Top announcements of the AWS Summit in New York, 2026, AWS News Blog
3. Amazon Web Services now introduces Context Ontology Accelerator, AWS What’s New
4. aws/context-ontology-accelerator, GitHub
5. AWS Glue Data Catalog now supports business context and semantic search (Preview), AWS What’s New
6. Adding business context, AWS Glue Developer Guide
7. AWS Glue Data Catalog now supports metadata exports to S3 Tables (Preview), AWS What’s New
8. Amazon S3 adds annotations to provide AI agents and analytics tools with context for data discovery, AWS What’s New
9. Knowledge graph, Amazon Quick User Guide
10. Amazon Quick desktop app is now generally available on macOS and Windows, AWS What’s New
11. Retrieve data and generate AI responses with Amazon Bedrock Knowledge Bases, Amazon Bedrock User Guide
12. Build a knowledge base by connecting to a structured data store, Amazon Bedrock User Guide
13. Amazon Bedrock Managed Knowledge Base is now generally available, AWS What’s New
14. Build a knowledge base with Amazon Neptune Analytics graphs, Amazon Bedrock User Guide
15. Amazon Bedrock Knowledge Bases supports GraphRAG now generally available, AWS What’s New
16. Amazon Bedrock AgentCore is now generally available, AWS What’s New
17. What is AWS Lake Formation?, AWS Lake Formation Developer Guide
18. Amazon Neptune, AWS
19. Catalog in IDC-based domains, Amazon SageMaker Unified Studio User Guide
20. Amazon SageMaker Catalog, AWS
21. Amazon SageMaker Data Agent integrates business context into conversations, AWS What’s New
22. Using the Lambda context object to retrieve Python Event Function information, AWS Lambda Developer Guide
23. Atlan integration, Amazon SageMaker Unified Studio User Guide
24. AWS SageMaker Unified Studio, Atlan Documentation
25. Set up Atlan MCP, Atlan Documentation
26. Build your context layer for AWS on Atlan, Atlan
27. Atlan - The Context Layer for AI, Atlan