RavenDB

RavenDB is a document database from RavenDB Ltd, which gives Hadera, Israel as its headquarters. It stores JSON documents without a fixed schema and keeps full ACID transactions, which many document stores relax. Queries are written in RQL, a SQL-style language, and the server builds and tunes indexes itself when it sees how data is queried. The vendor calls it the first database that optimises itself, a claim we cannot verify.

You can run it on your own servers (on-prem), as a managed service called RavenDB Cloud, or on constrained edge hardware, and the vendor says features are the same across all three. Recent additions are built-in vector search, embedding generation, GenAI tasks and an AI agent creator, plus time series, counters and attachments inside the same document. The vendor says about 12,000 customers use it. That figure is the vendor's.

• Automatic indexing. When a query arrives that no index serves, the server creates an auto-index and adapts it as query patterns change. Manually written static indexes remain available, including map-reduce and indexes over counters, time series and attachments.
• ACID transactions. Writes follow ACID rules per node, and cluster transactions extend them across a cluster. The vendor presents this as the difference from other NoSQL stores that treat consistency as an option.
• RQL query language. RQL looks like SQL and can query nested JSON, run full-text search and projections. Studio offers syntax validation and completion, and an in-Studio assistant helps with writing queries and troubleshooting.
• Multi-master clustering. Nodes accept writes and replicate, so work continues if a server fails. The default production cluster in the cloud has three nodes and can be extended by three more.
• Vector search and AI tasks. Semantic similarity search over vector embeddings runs inside the database, with embedding generation through external AI services. GenAI tasks handle summarisation, classification, tagging and enrichment from the management interface.
• Time series, counters and attachments. Timestamped data, incremental counters and large binary attachments live in the same document boundary as the JSON. Time series support rollups, retention rules and querying.
• ETL and subscriptions. Data can flow out through ETL tasks to SQL, Elasticsearch, Kafka, RabbitMQ, Snowflake and Azure Queue Storage, while data subscriptions let applications process changed documents in the background.

• A .NET product team shipping a line-of-business app. The team stores orders and customers as documents, lets auto-indexes handle new filters from the UI, and avoids writing migrations when fields change, using the Management Studio to inspect data.
• A device maker with intermittently connected equipment. An industrial tools company runs the edge build on resource-constrained hardware, stores readings as time series, and replicates to a central cluster when the network allows.
• A team adding semantic search to an existing catalogue. Developers enable embedding generation on product documents and query by vector similarity beside ordinary RQL filters, without moving data to a separate vector store.

There are three starting points. On the download page you choose release channel, operating system, architecture and version, obtain a free licence, then unzip the package and run it as the documentation describes. Alternatively, sign up on RavenDB Cloud, which offers free, developer and production tiers and is described as ready in minutes from a browser.

Client libraries and the documentation hub cover creating a client, connecting to the server and handling documents. The Community licence needs renewal once a year and the latest major version. A Developer licence is also free but limited to development, CI and QA use. Learning curve depends on RQL and index concepts, which the guides explain.

RavenDB Ltd gives its headquarters as Hadera, Israel. The Community licence is free for commercial use but limits a cluster to 3 cores and 6 GB RAM, needs the latest major version and renews yearly. It does not use MongoDB drivers.

RavenDB is entry #14 of 22 in the database management systems list: the products next to it are FerretDB above and ScyllaDB below. No numeric score is attached; established products follow the challengers in their ranked order.

Counting headquarters, 1 out of 22 sit in Middle East; for RavenDB that is Israel.

RavenDB Cloud runs on AWS, Google Cloud and Azure with a 99.99 percent uptime SLA stated on its page.

Offices: Hadera, Torun and New York.

Pricing as recorded: Free Community licence (3 cores, 6 GB RAM, commercial use allowed); Professional, Enterprise, AI and Cloud priced via sales or cloud portal.

Two licences are free: Community, valid for one year and eligible for commercial use, limited to 3 cores and 6 GB of RAM in a cluster of up to 3 nodes with unlimited databases, and Developer, free but not for commercial use, with 9 cores, 36 GB RAM and 3 nodes and a six-month renewal. Professional, Enterprise and AI licences are priced by core, with 6 GB of RAM per core on Professional and 12 GB on Enterprise and AI. Their amounts are not printed; the page offers contact forms and custom quotes. RavenDB Cloud is bought through the cloud portal. Yearly licences run 12 months, and orders in the last 30 days can be refunded on request. Higher values need a custom quote. OEM and ISV terms are available on request.

Read from ravendb.net when Best Database Management Systems in 2026 was last reviewed (7 November 2026).

The cloud page lists a 99.99 percent uptime SLA, SOC 2, ISO 27001, HIPAA and GDPR readiness, and X.509 and TLS 1.3 for authentication and encryption. The vendor says RavenDB Cloud runs on three major cloud providers, across 5 continents and more than 70 data centres, so you pick the region.

Encryption in transit and read-only certificates appear in the licence comparison. Self-hosted deployments follow your own hosting. Data export is possible through ETL tasks, subscriptions and backups, and the cloud API manages instances. A formal certificate register is not shown on the pages read.

Integrations the vendor names (12):

• Kafka
• RabbitMQ
• Elasticsearch
• Snowflake
• Azure Queue Storage
• Power BI
• Grafana plugin
• AWS
• Microsoft Azure
• SQL databases (ETL)
• OLAP ETL
• External AI embedding services

Closest by traits, not by rank: 4 products from the same list that overlap most with RavenDB.

• Redis (#6 of 22): Memory-first data store used for caching, queues and sessions, now tri-licensed. Watch out: the free database is limited to 30 MB on a shared deployment with community support.
• Couchbase (#8 of 22): Not for: buyers who want fully published self-managed prices up front. JSON document database with SQL++ queries, sold as Capella or as Server.
• FerretDB (#13 of 22): Open-source proxy that serves MongoDB drivers from PostgreSQL. Watch out: compatibility is partial and documented as known differences, so queries need testing.
• SurrealDB (#19 of 22): Multi-model database under BSL 1.1 with a free cloud instance. Suits: Teams wanting documents and graph relations in one engine. (United Kingdom)

Read from the vendor’s own pages on 5 November 2026.

• ravendb.net: positioning, AI features, deployment options, auto-indexing
• ravendb.net/why-ravendb: 12,000 customers claim, schema-less JSON, RQL, built-in features
• ravendb.net/buy: licence limits, Community and Developer terms, core and RAM rules, refund note
• ravendb.net/cloud: SLA, certifications, TLS 1.3, providers and data centres, machine sizes
• ravendb.net/download: download steps and free licences

We rank on what separates products after the demo: time to useful, the real price, the exit and who the product is for. The result for RavenDB: #14, out of 22 ranked. The editorial process sets out how facts are gathered and how commercial links are handled.