ArcadeDB

ArcadeDB is a multi-model database for teams that would otherwise run several databases side by side. Graphs, JSON documents, key-value pairs, vectors, BM25 full-text search, and time-series all live in one engine under the Apache 2.0 license. It works as an alternative to Neo4j and TigerGraph when you want graph features without an enterprise tier.

• Many query languages: SQL, OpenCypher, Gremlin, GraphQL, the MongoDB query language, and a native Java API all read the same data.
• Native vector search: embeddings are indexed with HNSW and queried alongside graph traversal in the same statement.
• Hybrid search: BM25 keyword ranking can be fused with vector similarity in a single query, then follow relationships from the top results.
• ACID transactions: a write-ahead journal rolls back uncommitted work after a crash, on one node or across several.
• High availability: multi-server replication is included, not sold separately.
• Flexible deployment: embed it inside a Java application or run it as a server with HTTP/JSON and several wire protocols.

Traversal speed stays constant no matter how large the database gets. A built-in graph analytics engine runs algorithms like PageRank and connected components, and the project's LDBC Graphalytics results show it ahead of Memgraph and FalkorDB on those tests.

Built-in importers move existing Neo4j and OrientDB databases over in minutes. Typical uses include knowledge graphs, GraphRAG and AI memory, fraud detection, recommendations, and customer 360 views.

Every feature is free. Arcade Data sells support subscriptions and consulting, and it has publicly committed to never changing the license.