Kendra
Amazon Kendra is closed to new customers. AWS put it into maintenance mode on 30 June 2026 and closed it to new customers on 30 July 2026 and its documentation now directs new work to Amazon Bedrock Knowledge Bases. Existing customers can continue to use it. A new AWS account cannot create a Kendra index.
The rest of this page describes what Kendra does, both for readers who already run it and to make clear what a replacement has to cover.
What Kendra does
Section titled “What Kendra does”Kendra is a managed information retrieval service. Rather than matching keywords, it uses semantic and contextual similarity to judge whether a passage answers a query, and ranks results accordingly. It handles three kinds of question:
- Factoid — who, what, when, where, answerable in a word or phrase, drawn from an FAQ or an indexed document
- Descriptive — answers that are a sentence, a passage or a whole document
- Keyword and natural language — short or ambiguous queries where the intended sense has to be inferred
Documents are ingested from connected repositories into an index. Document metadata drives filtering, so results can be scoped to a department, a document type or a security group.
Index editions
Section titled “Index editions”- GenAI Enterprise Edition — the most accurate, using current retrieval models; built for production and recommended by AWS where Kendra is used at all
- Basic Enterprise Edition — semantic search with high availability, suitable for production
- Basic Developer Edition — semantic search for proofs of concept, not recommended for production
Charging is per provisioned index, whether or not queries are run, plus charges for scanning and syncing documents through data source connectors.
Where it connected
Section titled “Where it connected”S3, databases, plain text repositories and third-party systems such as Microsoft SharePoint, through data source connectors. Access control is enforced at query time against the user or group, so search results respect the permissions of the underlying repository.
Kendra GenAI indices can be consumed from Amazon Q Business and Amazon Bedrock Knowledge Bases, and Kendra Intelligent Ranking can re-rank results produced by another search service.
What to use instead
Section titled “What to use instead”Amazon Bedrock Knowledge Bases is AWS’s stated replacement for retrieval-augmented generation and enterprise search. It handles chunking, embedding, vector storage and retrieval, and connects directly to the Bedrock models that consume the retrieved passages.
Amazon OpenSearch Service is the alternative where the requirement is search rather than generation, or where the retrieval pipeline needs to be assembled explicitly — it supports both lexical and vector search and gives full control over indexing, analysers and ranking.