SurrealDB

SurrealDB offers a single database engine that stores and queries documents, graphs, vectors, time-series and relational data within one ACID transaction boundary, together with built-in authentication and APIs. The stated design goal is to remove the need to stitch together separate vector databases, graph databases, document stores, caches and auth services, which the company frames as a source of data fragmentation, duplicated context, cumulative network latency and operational overhead.

The platform is presented as two entry points on one deployment: the database, where users define their own schema and query model, and Agent Memory, a memory layer for AI agents that sits on top of the database and is in early access. Agent Memory components include entity extraction, knowledge-graph construction, temporal facts and hybrid retrieval, and it is intended to assemble context from conversations, documents and connected systems. Architecturally, the product is described with distributed write nodes backed by S3 or S3-compatible object storage. Client libraries and type-safe graph traversals are cited for Rust, Go, TypeScript and Python, and a WebAssembly build is referenced for offline use.

Commonly cited use cases are GraphRAG (combining graph traversal with vector and full-text search in one query), knowledge graphs, persistent agent memory with per-user or per-tenant recall, and an organisation-wide queryable knowledge layer spanning documents, people, tools and decisions with permissions preserved. The company lists SOC 2 Type 2, ISO 27001, GDPR and Cyber Essentials Plus as compliance credentials.

Business model

SurrealDB provides a database engine and an associated agent memory layer used by engineering teams in production, alongside an official application, SurrealDB Studio. Open-source components are referenced, including the crud-bench benchmarking harness used to publish performance comparisons.

Traction

The company states the database runs in production at Samsung, Verizon, Tencent and PolyAI, and publishes case studies for PolyAI, Tencent and Later. Cited figures include 50 million graph edges across 8 million nodes at over 10,000 QPS on a single context graph at Tencent, approximately 30ms RAG latency at PolyAI, and Aspire scaling to 700,000 users within eight hours after replacing five backend tools with SurrealDB. Published benchmark comparisons of version 3.x against 2.x on identical hardware using crud-bench report a 31% mean CRUD improvement, 58% for batched operations, 136% for indexed read scans and 11894% for non-indexed full-table scans.

Latest developments

SurrealDB announced SurrealDB Studio, described as the new official app of SurrealDB, and released version 3.x with published benchmark gains over 2.x. The Agent Memory product is in early access with sign-ups open.