Machine Learning
Amazon Kendra
Amazon Kendra is an enterprise search service with connectors, relevance tuning, and document-level access controls. It can supply permission-aware passages to existing RAG applications over heterogeneous business repositories. It stopped accepting new customers on July 30, 2026.
Key points
- Connectors synchronize supported enterprise content sources
- User context and access-control lists can filter search results
- Query responses can include answers, documents, and passages
When to use it
- Ground an employee assistant in permissioned intranet content
- Search across enterprise repositories without building every crawler
Exam tips
- Treat Kendra as an existing-customer architecture; for new RAG designs, evaluate Bedrock Knowledge Bases or OpenSearch according to connector and retrieval-control needs
- Propagate user identity into every query or restricted documents can leak into prompts