Amazon Kendra pricing, AWS Kendra pricing and what a Kendra index costs per month

A single Amazon Kendra index costs between about $234 and $1,022 a month before add-ons, depending on the edition. GenAI Enterprise is $0.32 an hour, Basic Developer $1.125 and Basic Enterprise $1.40, billed every hour the index exists.

Most production indexes do not stay at the base size. Each base index includes one storage unit and one query unit, and anything above that is bought in more units. AWS's own GenAI example, 200,000 documents and 25,000 searches a day across four data sources, comes to $1,981.20 a month at 720 hours. Its Basic Enterprise example, 150,000 documents and 12,000 searches a day, comes to $2,021.28. Those examples are the honest middle of the range for a company search project, not the $234 headline.

Kendra is priced per index per hour, plus hourly storage and query units and a connector charge. You pay from the moment an index is created until it is deleted, whether or not anyone runs a search.

The base index includes one storage unit and one query unit. A storage unit adds 20,000 documents on GenAI Enterprise or 100,000 documents on Basic Enterprise. A query unit adds about 0.1 queries per second, roughly 8,000 searches a day, and is shared between the Query and Retrieve APIs. Basic Developer cannot add units at all, so it tops out at 10,000 documents and about 4,000 queries a day. Billing by the hour rather than by use is why forgotten test indexes are one of the most common sources of surprise Kendra charges.

For an index with two extra storage units, one extra query unit and three connectors, expect about $680 a month on GenAI Enterprise or about $2,571 a month on Basic Enterprise, at 730 hours.

The two editions are not interchangeable. GenAI Enterprise is far cheaper per hour but holds much less text per storage unit, so a document-heavy index can need many more units. The Basic Enterprise connector figure rests on our assumption of 30 minutes of sync a day per connector; heavier syncing raises it.

On GenAI Enterprise, connectors cost a flat $30 per index per month, which covers up to 500 hours of sync. On Basic Enterprise and Basic Developer, connectors cost $0.35 per hour while syncing plus $1 per million documents scanned.

The GenAI fee is per index, not per connector: AWS's own four-source example charges $30 in total. It is also charged in full for any month in which a connector is configured and is not prorated. On the Basic editions the sync meter is small for most companies, but a large first crawl of a SharePoint or file share estate can run for many hours. Connector usage is excluded from the free tier on every edition.

Yes, with limits. AWS gives up to 750 hours of free usage in the first 30 days on the Basic Developer or GenAI Enterprise edition. Basic Enterprise has no free tier, and connector usage never qualifies.

750 hours is roughly one index running for the whole month. Because Kendra only accepts existing customers since July 30, 2026, the free tier is now of practical use mainly to accounts that already had the service and want to test a second index before committing to it.

Not as a new customer. AWS put Amazon Kendra into Maintenance Mode effective June 30, 2026, and as of July 30, 2026 "the service is no longer open to new customers". Existing customers keep using it: AWS says the service "remains fully supported" with bug fixes and security updates, but new feature requests "will no longer be considered".

AWS's maintenance list defines the stage plainly: customers cannot onboard, and customers already using the service can continue. For a buyer that means two things. A team without Kendra today has to choose something else. A team with Kendra is paying for a service that will not gain features, which is usually the moment to price the replacement before the next budget cycle.

AWS points Kendra customers to Amazon Bedrock Managed Knowledge Base. Its own comparison lists 32+ connectors for Kendra against 7 for the Bedrock knowledge base.

That connector gap is the practical cost of the move. If your Kendra index pulls from Confluence, Jira, Salesforce, ServiceNow or a file share that the Bedrock knowledge base does not connect to, someone has to build and maintain that ingestion. Amazon Q Business, the employee-facing product many teams paired with Kendra, went into maintenance on the same date with Amazon Quick as its successor, so it is not the escape route either. We compare the realistic destinations, including self-built OpenSearch, on our Kendra alternative page.

On the infrastructure bill alone, often yes. A $680 a month GenAI index costs less than 50 people on any per-seat search product. The comparison changes once you add what Kendra does not include: a search interface, generated answers, identity mapping for permissions and the engineering time to run it.

Kendra returns ranked documents and passages. Turning those into a written, cited answer means calling a language model, for example through Amazon Bedrock, which is billed separately per token. Kendra supports document-level access control, but someone has to wire user identities and group membership into every query. If you have an engineer who already does that, Kendra was a fair deal. If you do not, and especially if you cannot buy Kendra at all, a finished product like InSearch is the shorter path, and you can try it on your own files with the tool at the top of this page.

Because they are old. A 2022 GovCloud reseller list still shows Enterprise Edition at $6.93 an hour with $3.465 storage and query units, figures from before AWS cut Kendra's unit sizes and prices in July 2021. The AWS Price List API also still carries legacy SKUs at $7 and $3.50 an hour.

Other pages are right but incomplete. A widely read 2023 cost article quotes $810 a month for Developer and $1,008 for Enterprise, which matches AWS at 720 hours, but it predates the GenAI Enterprise edition and the maintenance notice. When you check a quote, match it against aws.amazon.com/kendra/pricing for your region and count hours the way AWS does: from creation to deletion.