SurrealDB

SurrealDB is a multi-model database for documents, graphs, vectors and relational data, built for AI and app developers, run in its cloud or self-hosted.

SurrealDB keeps documents, graph relationships, vectors, time-series and relational records in one engine, so an application can query them together in a single ACID transaction. That removes the work of running a document store, a graph database and a vector index side by side and keeping them in step. It is built by SurrealDB Ltd, a London company founded by Tobie and Jaime Morgan Hitchcock. The database launched in 2022. Queries are written in SurrealQL, one language covering relational selects, graph traversal, vector and full-text search, temporal and key-value lookups. Tables can be schemaless and accept records of different shapes, or schemafull with typed fields defined up front. Vector fields take HNSW indexes for nearest-neighbour search, and full-text indexes rank matches with BM25 and return highlights. Live queries push each committed write to subscribed clients, while change feeds let another service replay a table's history from a chosen point. Permissions sit beside the schema at the role, record and field level, so access is enforced inside the database. SurrealDB suits developers building retrieval and agent features who would otherwise wire a vector store to a graph database. It also fits product teams that want live updates in the client, and SaaS builders who keep per-user access rules in the data layer. The engine can run embedded in Python and JavaScript runtimes, through WebAssembly in the browser, or at the edge. Production workloads that need high availability run as a multi-node cluster on SurrealDB Cloud, and regulated buyers can self-host the Enterprise edition with FIPS and audit controls. The engine itself is source-available under the Business Source License 1.1. Typical jobs include GraphRAG retrieval that joins graph traversal with vector search, knowledge graphs of entities and relationships, and digital twins that tie assets to their telemetry. SurrealDB Agent Memory runs on the same engine and gives agents persistent, queryable memory with entity extraction, temporal facts and hybrid retrieval. Through the MCP server that ships inside the engine, an AI assistant can query a self-hosted or local database directly. For an application that reads the same data as documents, as a graph and as vectors, SurrealDB holds all three under one query language and one set of permissions.