OpenSearch
AWS's Elasticsearch fork. Apache 2.0, AWS-managed via OpenSearch Service.
VISIT OPENSEARCHKey takeaway: The deciding factor is almost never the engine, it is where you already are. If your infrastructure lives in AWS or your legal team requires Apache 2.0, OpenSearch is the answer and the comparison with Elasticsearch stops there. Outside those two conditions, Elasticsearch is the more central project with the larger ecosystem behind it.
Quick facts
- CategoryDatabase-native FTS
- EngineJava
- PricingOpen source
- LicenseApache-2.0
- Created2021
- GitHub stars11k
- Vector supportYes
- Edge-readyNo
What it is
OpenSearch is AWS's 2021 fork of Elasticsearch + Kibana, kept Apache-licensed throughout. Functionally near-identical to Elasticsearch with AWS-aligned tooling and the AWS-managed OpenSearch Service. Used heavily in AWS-locked organisations.
Best for
- AWS-locked organisations needing managed search
- Workloads requiring strict Apache 2.0 licensing
- Teams already using AWS-native services who want one vendor
When not to pick it
Outside AWS the alignment is the differentiator. New non-AWS projects almost always pick Elasticsearch (now AGPL again) or one of the lighter options.
My take
AWS-aligned organisations should use OpenSearch. Outside that audience Elasticsearch is the more central project.
The fork has diverged more than people expect
OpenSearch began as a 2021 fork of Elasticsearch and Kibana, but several years of separate development mean the two are no longer interchangeable in practice. Client libraries are distinct packages and the official Elastic clients refuse to connect to OpenSearch. Vector search, machine-learning inference and anomaly detection were built independently on each side and expose different APIs. OpenSearch Dashboards has drifted from Kibana in both features and plugin compatibility. Governance moved to the Linux Foundation in 2024, which made the project less obviously an AWS artefact and more of a neutral one, and that has helped adoption outside AWS shops. The practical advice: treat any tutorial written for Elasticsearch as approximately correct rather than directly applicable, and check the OpenSearch documentation for anything beyond basic indexing and querying.
Running it on AWS, and what managed really means
Amazon OpenSearch Service is why most teams are here: cluster provisioning, patching, snapshots and IAM-based access without operating JVMs yourself. Managed is not the same as free of operations. You still choose instance types and shard counts, you still plan version upgrades, and a badly sized cluster still falls over halfway through a reindex. The serverless option removes capacity planning in exchange for a different cost model and a narrower feature set, so read the limitations before assuming it is a drop-in. Cost scales with provisioned capacity rather than query volume, which punishes idle clusters and rewards steady load, the opposite curve to a per-request product like Algolia. If you are standing up one small index because OpenSearch Service is the familiar button, price a Typesense or Meilisearch instance against it first.
Frequently asked questions
Is OpenSearch the same as Elasticsearch?
It started as a fork of it and is still recognisably similar, but they have diverged. Clients are separate and incompatible, vector search and machine-learning features were developed independently with different APIs, and Dashboards no longer tracks Kibana. Basic indexing and querying transfer across; anything more advanced needs the OpenSearch documentation rather than an Elasticsearch tutorial.
Is OpenSearch free?
The software is Apache 2.0 and free to self-host, with no licence ambiguity, which is one of its main selling points. Amazon OpenSearch Service is the paid managed offering and bills on provisioned capacity rather than per query, so an idle cluster still costs you. Self-hosting is free in licence terms and expensive in operations terms.
OpenSearch or Elasticsearch for a new project?
OpenSearch if you are on AWS and want the managed service, or if Apache 2.0 is a hard requirement. Elasticsearch otherwise, since it has the bigger ecosystem, more documentation, more third-party integrations and faster feature movement. If neither AWS nor licensing is decisive, the honest answer is that you may not need either one.
Links
Compare OpenSearch side-by-side
Similar tools you should also consider
If OpenSearch is your pick, the next conversation is short
The 30-min call is where your search choice becomes a real architecture, a relevance-tuning plan, and a price range you can take to your stakeholders.