Skip to main content
XenonStack

Artificial Intelligence

XenonStack Blog Amazon Kendra Implementation Guide: Going into 2027

and | Published | Updated

Amazon Kendra Implementation Guide: Going into 2027
13:43

Key takeaway

  • Amazon Kendra is in maintenance mode and closed to new customers. See how an S3 index works, the cost and governance points, and when to move to Bedrock.

Amazon Kendra is a managed intelligent search service, but it is no longer a safe default for a new build. AWS states that it put Kendra into maintenance mode from 30 June 2026 and that the service is no longer open to new customers from 30 July 2026. Existing customers keep bug fixes and security updates, and AWS recommends Amazon Bedrock Managed Knowledge Base for new search applications (AWS availability change notice, checked on 12 November 2026).

This guide is for teams that already run, or inherit, a Kendra index. It explains what Kendra does, how an S3-based implementation is put together, what the governance and cost points are, and how to decide between staying and moving. For the product overview see Amazon Kendra benefits for businesses, and for retrieval with agents see Amazon Kendra with AI agents.

What is the status of Amazon Kendra in 2026?

We read the AWS documentation on 12 November 2026 for service changes, retirement notes and any mention of the GenAI Index. The table records what AWS actually says.

QuestionWhat the AWS documentation saysSource
Is Kendra still supported?Yes. The service remains fully supported, with bug fixes and security updates for existing customers. New feature requests are no longer considered.Availability change
Can new customers start?No. The overview gives 30 July 2026 as the date from which the service is no longer open to new customers, and the page banner says the same without a date.Availability change
Is there a retirement date?No end-of-support date is given on the availability page.Availability change
What does AWS recommend instead?Amazon Bedrock Managed Knowledge Base, with a Retrieve and Generate API and an Agentic Retrieval API. The banner also points to Amazon Bedrock Knowledge Bases.Availability change
What about the GenAI Index?The availability page does not mention it. The Kendra overview still lists GenAI Enterprise Edition as an index type, described as suited to retrieval augmented generation (RAG) and usable with Amazon Q Business and Amazon Bedrock knowledge bases.What is Kendra
What about Amazon Q Business?Its documentation says it is also no longer open to new customers and points to Amazon Quick for similar capabilities.Q Business overview

The banner and the overview on the availability page are worded differently, so confirm the position for your own account with AWS Support before you commit budget either way.

What does Amazon Kendra do and which questions can it answer?

Amazon Kendra is a managed information retrieval service that uses natural language processing and deep learning to judge whether a passage or document is relevant to a query, instead of matching keywords alone. AWS lists three query types it supports: factoid questions (who, what, when, where), descriptive questions that need a passage or longer answer, and keyword or natural language queries where it infers the intended meaning. Factoid answers can come from a FAQ or from indexed documents.

Which Kendra index types exist?

Index typeWhat AWS saysWhen it fits
GenAI Enterprise EditionHighest accuracy, high availability, built for productionProduction RAG and enterprise search on an existing estate
Basic Enterprise EditionSemantic search with high availability, suitable for productionProduction search without generative AI needs
Basic Developer EditionSemantic search for proof of concept, not recommended for productionTesting only

Source: AWS, What is Amazon Kendra. AWS also notes that search results can reflect your organisation's security model and be filtered by user or group access.

How does an Amazon Kendra implementation work?

A Kendra implementation has four parts: documents held in a source repository, a data source connector that syncs them, an index that serves search, and a client application that calls the search API. For many teams the source is an Amazon S3 bucket.

What do you need before you start?

  • An AWS account with permission to create the resources you need.
  • An Amazon S3 bucket in the same region as the index, holding the documents or FAQ content.
  • An IAM role the connector uses to read the bucket. AWS says the console can create it, and that a role used for an index cannot be reused for a data source.
  • A decision on access control: supply an access control list file in JSON if users should see only the documents they are allowed to open.

Source: AWS, Amazon S3 data source connector.

What is the setup sequence?

  1. Create the bucket and upload the documents.
  2. Create the index and choose the index type. Use a developer index only for testing.
  3. Add an S3 data source, point it at the bucket, attach the IAM role and set the sync schedule.
  4. Run the first sync. The first sync crawls everything.
  5. Optionally add FAQ content, then test queries in the search console.
  6. Call the search API from your website, chatbot or agent-assist tool.

Console labels change between releases, so follow the current AWS guide for each click. Two connector details matter in practice: the S3 connector supports full syncs and syncs of new, modified and deleted files only, and AWS sets a 50 MB file size limit for the connector. Metadata can sit in a separate metadata folder, one file per document, and is mapped to index fields.

Should you stay on Kendra or move to a Bedrock knowledge base?

AWS publishes a migration guide on the availability page. It says to run Kendra and the managed knowledge base in parallel and compare relevance (NDCG or MRR), latency, throughput and completeness. The table lists differences it names.

Two-column diagram. Left: Kendra connectors, search features, learning and enrichment, index type. Right: connector coverage, feature gaps, parallel test, index cost. Outcome: stay, migrate or retire.
AreaKendraBedrock Managed Knowledge Base
Connectors32+ native connectors7: Amazon S3, Confluence, Microsoft SharePoint, Web Crawler, Google Drive, Microsoft OneDrive and a custom connector
Search modeSemantic search availableAlways hybrid (keyword plus semantic); no semantic-only mode
ChunkingNot specified in the AWS guideNo semantic chunking; AWS suggests fixed-size, 200 tokens, 30% overlap as a start
Query suggestions, faceted search, custom synonyms, spell checkingBuilt inNo direct equivalent; AWS lists application-side workarounds
Incremental learning, document enrichmentBuilt inNo direct equivalent; AWS suggests external feedback signals and a Lambda or Step Functions pipeline

Source: AWS availability change and migration guide. A simple rule: stay on Kendra while an existing index meets its relevance targets and relies on features with no direct equivalent, and plan the move for any workload that depends on connectors outside the seven above or that you are about to extend.

What are the governance, security and cost points?

  • Access: results can be filtered by user or group, but customers are responsible for authenticating and authorising users.
  • Cost: after the free trial (up to 750 hours in the first 30 days on a GenAI Enterprise Edition or Developer Edition index), AWS charges for all provisioned indexes, even empty ones with no queries, plus data source scanning and syncing. Connector usage does not qualify for the free tier. Delete unused test indexes.
  • Evidence: keep a record of which documents each connector syncs and who can see them, so an answer can be traced back to its source.

Source: AWS, What is Amazon Kendra. For agents that act on retrieved content, govern the retrieval layer as well as the model. ElixirData Context OS keeps governed context, versioned policy and decision traces in one decision layer above AI agents, and AgentSearch offers context-aware enterprise search across unstructured knowledge bases. See also our guide to AI agents and our knowledge graph services.

What are the trade-offs, and when should you not use Kendra?

  • Do not choose Kendra for a new programme, because new customers can no longer sign up.
  • Do not leave a developer index running as a production service, because AWS does not recommend it.
  • Do not assume a move is a lift and shift: connectors, filters and metadata rules differ, as the table above shows.
  • Do not plan around a retirement date, because AWS has not published one. Plan around the maintenance mode position instead.

What adoption sequence is sensible for an existing Kendra estate?

  1. List every index, data source and the applications that call it.
  2. Record which Kendra-only features each application uses.
  3. Check connector coverage against the seven in the managed knowledge base, and plan an export to S3 for the rest, which is the pattern AWS recommends.
  4. Build a test set of real questions and run both services in parallel.
  5. Decide per application: stay, migrate, or retire.
  6. Document the decision and review it when AWS updates the notice.

If you run agents on top of retrieval, ElixirClaw is an agentic OS platform to build, orchestrate and govern AI agents and automation workflows while keeping humans in control.

To map one search or retrieval decision your agents should make, and the context and policy it depends on, book a Context Workshop. See what happens in the workshop.

What should you plan for in 2027?

Plan around the position the guide describes: Kendra is in maintenance mode, existing indexes keep working, and no retirement date has been published. These points keep an existing estate under control.

  • Know what you run. List every index, data source and calling application, and record which Kendra only features each one uses.
  • Decide per application. Stay, migrate or retire, based on a parallel test of real questions against the Bedrock managed knowledge base.
  • Control cost and access. Delete unused test indexes, keep the record of what each connector syncs and who can see it, and review the decision whenever AWS updates its notice.

Frequently asked questions about Amazon Kendra

Is Amazon Kendra still available?

Existing customers can keep using it. AWS put Kendra into maintenance mode from 30 June 2026 and states it is no longer open to new customers from 30 July 2026. Bug fixes and security updates continue.

Has AWS announced a retirement date for Amazon Kendra?

The AWS availability page gives no end-of-support date. It describes maintenance mode, with bug fixes and security updates for existing customers.

What does AWS recommend instead of Amazon Kendra?

AWS recommends Amazon Bedrock Managed Knowledge Base for new search applications and for migrating existing Kendra applications.

What is the Amazon Kendra GenAI Index?

AWS lists GenAI Enterprise Edition as an index type with the highest accuracy, suited to retrieval augmented generation and usable with Amazon Q Business and Amazon Bedrock knowledge bases. The availability change page does not mention it.

How does Amazon Kendra connect to documents in Amazon S3?

Through the S3 data source connector. The bucket must be in the same region as the index, the connector uses an IAM role, and it supports full syncs and syncs of new, modified and deleted files.

How is Amazon Kendra priced?

After a free trial of up to 750 hours in the first 30 days on eligible indexes, AWS charges for every provisioned index, even an empty one, plus data source scanning and syncing. Connector usage does not qualify for the free tier.

Next Steps in Optimizing Your Search Experience with Amazon Kendra

Talk to our experts about implementing a compound AI system with Amazon Kendra. Learn how industries and different departments use Agentic Workflows and Decision Intelligence to become decision-centric. Amazon Kendra utilizes AI to automate and optimize enterprise search, IT support, and operations, improving efficiency and responsiveness.

Follow XenonStack on LinkedIn: XenonStack on LinkedIn for new guides on agentic AI, data and platform engineering.

Table of Contents

navdeep-singh-gill

Navdeep Singh Gill

Global CEO and Founder of XenonStack

LinkedIn profile

Navdeep Singh Gill is serving as Chief Executive Officer and Product Architect at XenonStack. He holds expertise in building SaaS Platform for Decentralised Big Data management and Governance, AI Marketplace for Operationalising and Scaling. His incredible experience in AI Technologies and Big Data Engineering thrills him to write about different use cases and its approach to solutions.

Get the latest articles in your inbox

Subscribe Now

Interested in solving your challenges with the XenonStack team?
Get Started

Tell us about you and your priorities so we can prepare the right conversation. It takes about 2 minutes.

Please Select your Industry
Banking
Fintech
Payment Providers
Wealth Management
Discrete Manufacturing
Semiconductor
Machinery Manufacturing / Automation
Appliances / Electrical / Electronics
Elevator Manufacturing
Defense & Space Manufacturing
Computers & Electronics / Industrial Machinery
Motor Vehicle Manufacturing
Food and Beverages
Distillery & Wines
Beverages
Shipping
Logistics
Mobility (EV / Public Transport)
Energy & Utilities
Hospitality
Digital Gaming Platforms
SportsTech with AI
Public Safety - Explosives
Public Safety - Firefighting
Public Safety - Surveillance
Public Safety - Others
Media Platforms
City Operations
Airlines & Aviation
Defense Warfare & Drones
Robotics Engineering
Drones Manufacturing
AI Labs for Colleges
AI MSP / Quantum / AGI Institutes
Retail Apparel and Fashion

Continue