Agentic AI

AWS Context – Knowledge Graph for AI Agents

~ Last updated on : ~ Kiro Agent ~ Leave a comment

AWS Context Overview

  • AWS Context is a new service announced at AWS Summit New York City (June 17, 2026) that automatically builds a knowledge graph from your existing organizational data so AI agents can find the right information, provide correct answers, and take the right actions.
  • AWS Context maps the relationships across existing data into a knowledge graph and provides agentic search so AI agents can access governed data relationships, business rules, and domain knowledge at runtime.
  • It eliminates the need to build custom retrieval pipelines, provision infrastructure, or manually wire agents to individual data sources.
  • AWS Context is currently in “Coming Soon” status (as of June 2026).
  • The service is built on the same knowledge graph technology that powers Amazon Quick (formerly Amazon Q), where hundreds of thousands of users interact daily with a production knowledge graph processing millions of requests per day.

Key Features

Automatic Relationship Mapping

  • Automatically infers relationships between data assets, business rules, and domain knowledge across the organization.
  • Understands what tables exist, what’s stored in different columns, which sources are the most authoritative, and how they relate to each other.
  • Data stewards and curators manage the graph through an intuitive console experience, reviewing inferred relationships, promoting them to production, and attaching domain-specific knowledge.

Broad Data Source Connectivity

  • Connects to all organizational data including:
    • Databases (relational, NoSQL, data warehouses)
    • Slack messages and team communications
    • Documents and wikis
    • Emails
    • CRM systems
    • Data lakes, data warehouses, and lakehouses
    • Data streams
  • Designed to connect to third-party catalogs, so context from systems beyond AWS can be brought into the same graph.

Context That Learns (Continuous Learning Loop)

  • AWS Context gets smarter the more agents use it.
  • As agents query the graph, it observes:
    • Which sources produce correct results
    • Which join paths agents rely on
    • Which curated rules get applied
  • Ranks sources by actual usage and shares learnings across the organization.
  • When one agent discovers a correct join path or resolves a schema ambiguity, other agents pick it up automatically without requiring human re-curation.
  • Every agent improves based on the findings of a single query.

Open and Portable by Design

  • All key metadata from structured and unstructured sources is published into Apache Iceberg format in Amazon S3 Tables.
  • Context can be queried with Amazon Athena, Amazon Redshift, Apache Spark, or any Iceberg-compatible engine.
  • Build downstream systems on it, audit it, or migrate it — your context stays fully yours.
  • Agents query it through agentic search APIs and MCP tools, whether built on Amazon Bedrock AgentCore, deployed on Amazon EKS, or running on MCP-compatible frameworks.

Identity-Aware Governance

  • Every query is identity-aware — each call inherits the calling user’s IAM and Lake Formation permissions.
  • An agent can only see and traverse the relationships its identity is authorized to access.
  • Every interaction is auditable — security and compliance teams can verify what an agent accessed and under what authority.
  • Uses the same access controls organizations already rely on (IAM, Lake Formation).

Zero Infrastructure Management

  • No infrastructure to provision — fully managed service.
  • No retrieval pipeline to build — agents navigate the knowledge graph directly.
  • Begin gathering and curating context with just a few clicks in the AWS Management Console.

Architecture & How It Works

  • Knowledge Graph Foundation: Built on the same technology that powers Amazon Quick’s production knowledge graph (catalogs datasets, dashboards, and metadata at scale).
  • From Personal to Organizational: Extends what was a personal knowledge graph (Amazon Quick) into an organizational one — a shared, governed context layer for all agents and applications.
  • Integration Points:
    • AWS Glue Data Catalog
    • Amazon SageMaker Unified Studio
    • AWS Lake Formation
    • Amazon Bedrock AgentCore
    • Amazon Bedrock Managed Knowledge Base
  • Agent Access: Agents query through agentic search APIs and MCP tools — framework agnostic.
  • Data Flow:
    1. AWS Context connects to organizational data sources (databases, documents, Slack, CRMs, etc.)
    2. Automatically maps relationships and infers context
    3. Data stewards review, promote, and curate inferred relationships via console
    4. Metadata published in Iceberg format to S3 Tables
    5. Agents query the graph at runtime with identity-aware permissions
    6. Learning loop continuously improves source ranking and path resolution

Integration with AWS Services

  • Amazon Quick: When AWS Context is enabled, Quick’s agents gain access to the broader enterprise knowledge graph, including cross-system relationships, business rules, and curated context beyond any single user’s personal graph.
  • AWS Glue Data Catalog: Integrates with the knowledge graph; supports new business context, semantic search, and skill assets (preview).
  • Amazon Bedrock Managed Knowledge Base: Plugs into AWS Context to enable agentic search across all structured, unstructured, and domain data.
  • AWS Lake Formation: Provides permission governance layer — agents inherit Lake Formation permissions.
  • Amazon S3 Tables: Stores all metadata in Iceberg format for open, queryable access.
  • Amazon S3 Annotations (GA): Attach rich, queryable business context directly to S3 objects — up to 1 GB of context per object, mutable, and automatically queryable through S3 Metadata.

AWS Context vs. Bedrock Knowledge Bases vs. Neptune vs. Glue Data Catalog

Feature AWS Context Bedrock Knowledge Bases Amazon Neptune AWS Glue Data Catalog
Primary Purpose Organizational knowledge graph for AI agents RAG over unstructured documents General-purpose graph database Metadata catalog for data assets
Data Type Structured + unstructured + institutional knowledge Primarily unstructured (documents, PDFs, web pages) Structured graph data (nodes, edges, properties) Technical metadata (schemas, tables, partitions)
Relationship Handling Automatically infers and learns relationships No explicit relationships — vector similarity only Manually defined graph relationships (RDF/Property Graph) Catalog lineage only — no semantic relationships
Learning/Improvement Continuous learning from agent usage patterns No learning — static retrieval pipeline No learning — requires manual graph updates No learning — manual catalog maintenance
Governance Identity-aware (IAM + Lake Formation per query) Basic access control on knowledge base level IAM-based cluster access Lake Formation fine-grained access
Infrastructure Fully managed — no provisioning needed Managed — requires data ingestion setup Self-managed clusters or serverless (must provision) Managed catalog service
Agent Integration Native agentic search APIs + MCP tools Integrated with Bedrock agents via RAG retrieval Custom integration via query APIs (Gremlin/SPARQL) API-based catalog lookup
Data Sources Databases, Slack, emails, CRMs, documents, streams S3, SharePoint, Confluence, Google Drive, web crawlers Application-loaded graph data AWS data service schemas (S3, RDS, Redshift, etc.)
Metadata Format Apache Iceberg in S3 Tables (open, portable) Vector embeddings in managed/custom vector stores Property Graph / RDF triples Hive-compatible catalog format
Best For Enterprise agents needing cross-system business context QA over document collections (policies, manuals, docs) Complex graph traversal, fraud detection, social networks ETL pipeline management, schema discovery
Use with AI Agents Purpose-built for agents — agents navigate graph directly Agents retrieve relevant chunks via similarity search Agents query graph via custom code Agents discover table metadata only

When to Use Which Service

  • AWS Context: Use when you need agents to understand business relationships across multiple systems — understanding how customer data in your CRM relates to orders in your database and communications in Slack.
  • Bedrock Knowledge Bases: Use when agents need to answer questions from unstructured document collections (policy documents, product manuals, knowledge bases) via RAG.
  • Amazon Neptune: Use when you have complex, explicitly defined graph relationships requiring traversal queries — fraud detection rings, social networks, recommendation engines.
  • AWS Glue Data Catalog: Use for ETL pipeline management, schema discovery, and technical metadata governance across your data lake.
  • Combined Approach: AWS Context integrates with Bedrock Managed Knowledge Base to provide agentic search across all structured, unstructured, and domain data together.

Use Cases

Customer Support Agents

  • A customer support agent triaging an issue needs to pull up purchase history, shipping status, and return eligibility across multiple different sources.
  • With AWS Context, the agent navigates the knowledge graph to find all relevant data without custom integrations per data source.
  • The next time a similar issue arises, the agent knows exactly where to go, reducing resolution time.

Data Analyst Agents

  • Agents can discover authoritative data sources, understand join paths between tables, and know which filters and aggregation rules apply.
  • Business rules (like “always exclude test accounts from revenue calculations”) are captured in the knowledge graph and applied automatically.
  • Reduces time spent searching for the right data and understanding how to use it correctly.

Compliance & Audit Agents

  • Compliance agents can trace data lineage and access patterns across the organization.
  • Every agent interaction is auditable — security teams can verify what was accessed and under whose authority.
  • Identity-aware governance ensures agents only access data they’re authorized to see, maintaining regulatory compliance.

Sales & CRM Agents

  • Agents can see the latest interactions with a customer in the CRM and recommend the best follow-up actions.
  • Cross-referencing emails, Slack conversations, and deal history provides complete customer context.
  • Without context, agents confidently give recommendations that are wrong — AWS Context solves this.

Enterprise Knowledge Management

  • Captures institutional knowledge that has never been written down — business rules, domain expertise, tribal knowledge.
  • Makes organizational wisdom available to every agent, not just the humans who happen to know it.
  • New agents benefit immediately from the accumulated context of the entire organization.

Key Benefits

  • Faster Time to Value: No retrieval pipeline to build, no infrastructure to provision — start with a few clicks.
  • Compounding Intelligence: Gets smarter with every agent interaction across the organization.
  • Reduced Token Consumption: Agents navigate directly to the right information instead of processing large context windows.
  • Enterprise Governance: Built-in identity-aware access control using existing IAM and Lake Formation policies.
  • Open Standards: Iceberg format means no vendor lock-in for metadata — query with any compatible tool.
  • Cross-Agent Learning: One agent’s discovery benefits all agents in the organization.
  • Framework Agnostic: Works with Bedrock AgentCore, EKS-deployed agents, or any MCP-compatible framework.

AWS Certification Exam Practice Questions

Question 1: A company wants to enable its AI agents to access business context from multiple data sources including databases, Slack messages, and CRM systems, with automatic relationship inference and identity-aware governance. The solution should require no infrastructure provisioning. Which AWS service should they use?

  1. Amazon Neptune with GraphRAG
  2. AWS Context
  3. Amazon Bedrock Knowledge Bases
  4. AWS Glue Data Catalog with Lake Formation
Show Answer

Answer: B –

Explanation: AWS Context automatically builds a knowledge graph from existing organizational data (databases, Slack, CRMs, documents, emails), infers relationships, provides identity-aware governance, and requires no infrastructure provisioning. Neptune requires cluster management, Bedrock Knowledge Bases focus on unstructured RAG, and Glue Data Catalog only manages technical metadata.

Question 2: How does AWS Context store its metadata to ensure portability and avoid vendor lock-in?

  1. In Amazon DynamoDB tables with proprietary format
  2. In Amazon Neptune graph database clusters
  3. In Apache Iceberg format in Amazon S3 Tables
  4. In Amazon OpenSearch vector indexes
Show Answer

Answer: C –

Explanation: AWS Context publishes all key metadata from structured and unstructured sources into Apache Iceberg format in Amazon S3 Tables. This open format allows customers to query context with Amazon Athena, Amazon Redshift, Apache Spark, or any Iceberg-compatible engine, ensuring portability and no vendor lock-in.

Question 3: A customer support agent built with AWS Context discovered the correct join path between order data and shipping status. What happens when another agent in the organization faces a similar query?

  1. The other agent must independently discover the same join path
  2. A data engineer must manually configure the path for the other agent
  3. The other agent automatically benefits from the discovered path through the learning loop
  4. The organization must rebuild the knowledge graph to include the new path
Show Answer

Answer: C –

Explanation: AWS Context features a continuous learning loop. When one agent discovers a correct join path or resolves a schema ambiguity, it ranks sources by actual usage and shares what it learns across the organization. Other agents automatically pick up these discoveries without requiring human re-curation.

Question 4: Which of the following statements about AWS Context governance are correct? (Select TWO)

  1. Each query inherits the calling user’s IAM and Lake Formation permissions
  2. All agents have unrestricted access to the entire knowledge graph
  3. Every agent interaction is auditable by security and compliance teams
  4. Governance rules must be configured separately from existing AWS permissions
  5. Access control only applies at the knowledge graph level, not per-query
Show Answer

Answer: A, C

Explanation: AWS Context makes every query identity-aware. Each call inherits the calling user’s IAM and Lake Formation permissions, so an agent can only see relationships its identity is authorized to access. Because access runs through identity, every interaction is auditable — security teams can verify exactly what was accessed and under what authority.

Question 5: A company needs to ground its AI agents in both structured business relationships (from databases and CRMs) AND unstructured documents (policy manuals, product guides). Which approach provides the most comprehensive solution?

  1. Use Amazon Neptune for all data types
  2. Use AWS Context alone for both structured and unstructured data
  3. Use AWS Context integrated with Amazon Bedrock Managed Knowledge Base
  4. Use AWS Glue Data Catalog with Amazon Bedrock Knowledge Bases
Show Answer

Answer: C –

Explanation: AWS Context integrates with Amazon Bedrock Managed Knowledge Base to enable agentic search across all structured, unstructured, and domain data. AWS Context provides the knowledge graph for structured relationships and business rules, while Bedrock Managed Knowledge Base handles unstructured document retrieval — together they provide comprehensive coverage.

Frequently Asked Questions

What is AWS Context?

AWS Context is a service that automatically builds a knowledge graph from your organizational data — databases, documents, Slack messages, CRMs, emails. It infers relationships between data assets and makes them navigable by AI agents with built-in governance controls.

How does AWS Context differ from Bedrock Knowledge Bases?

Bedrock Knowledge Bases provide RAG over unstructured documents (PDFs, web pages). AWS Context builds a structured knowledge graph that understands relationships between entities, business rules, and data lineage across all your systems — giving agents navigational intelligence, not just text retrieval.

Does AWS Context require infrastructure setup?

No. AWS Context is fully managed with no infrastructure to provision and no retrieval pipeline to build. It stores metadata in Iceberg format in S3 Tables and learns continuously from agent interactions to improve accuracy over time.

References