Skip to content
NEW BLOG

Using Surrealism to build your own extensions

Read blog

1/2

SurrealDB hero illustration

Context, made atomic

Documents, graphs, vectors, time-series, and relational data as native primitives in a single ACID transaction. No plugins, no bolt-ons, no frankenstack.

  1. Babcock
  2. ING
  3. Nvidia
  4. Samsung
  5. Tencent
  6. PolyAI
  7. Verizon
  8. Walmart
  9. Babcock
  10. ING
  11. Nvidia
  12. Samsung
  13. Tencent
  14. PolyAI
  15. Verizon
  16. Walmart

Native multi-model engine

Documents, graphs, vectors, full-text search, time-series, and relational data - fully unified in one database and query language. No plugins or bolt-ons.

Result mode view

AI data layer

Built for context, memory, search, and retrieval - combining vectors, graphs, full-text search, and relational joins in one engine.

AI data layer illustration

Deploy anywhere, scale horizontally

Same API runs from in-memory, embedded and edge devices to self-hosted or highly scalable in the cloud.

Deploy anywhere illustration

Real-time and event-driven

Built-in real-time subscriptions, event triggers, and streaming updates to power reactive, real-time experiences.

Real-time messaging illustration

Enterprise security and compliance

RBAC, record-level permissions, JWT auth, multi-tenant isolation, and SOC 2 / ISO 27001 compliance built in.

Authentication and users illustration

Developer tools

Surrealist visual IDE, WebAssembly plugins, SurrealMCP for agent-native access, and SurrealQL - the best of SQL, NoSQL, and graph in one syntax.

SurrealQL query pane

High-performance, reliable core

Built in Rust with full ACID guarantees. Memory-safe, thread-safe, and crash-resilient by design.

Rust-powered performance illustration
See all features

HOW IT WORKS

Multi-model query architecture

One query flows through documents, graphs, vectors, and time-series - resolved from a single transactional snapshot.

SurrealDB SDK
Postgres-compatible client
GraphQL client
SurrealDB
SurrealDB
RPC protocol
PostgreSQL wire protocol
GraphQL endpoint
SurrealQL parser
ANSI-SQL parser
GraphQL parser and schema generator
SurrealDB executor
Execution logical plan
Execution physical plan
Document processing
SurrealDS
Key-value engine data storage and retrieval

DEPLOY ANYWHERE

Single binary, any device

In-memory, embedded, edge, WASM, or cloud - the same engine runs everywhere.

Python
Python, C, and C++ execution environment
In-process native communication
SurrealDB
SurrealDB
Query engine
Storage engine(Local disk or block storage)
Periodic storage data compaction
and compacted data storage
SurrealDS
Local disk

HOW SURREALDB COMPARES

Every other agent stack is a frankenstack

Document store, graph database, vector database, cache, message broker, memory middleware - glued together and prayed over. SurrealDB replaces all of it.

vs. Postgres

Postgres bolts on vectors with pgvector and graphs with recursive CTEs. SurrealDB makes every model native - no extensions, no workarounds.

vs. MongoDB

MongoDB is a document store. Graph traversal, vector search, and temporal queries require separate systems. SurrealDB unifies them.

vs. Neo4j

Neo4j is a graph database. Documents, vectors, and time-series require separate infrastructure. SurrealDB does it all in one engine.

vs. Vector databases

Pinecone, Chroma, and Weaviate store embeddings. SurrealDB stores embeddings alongside their entities, relationships, and temporal context.

vs. Memory middleware

Mem0 layers memory on top of fragmented databases. Spectron builds memory directly into SurrealDB - one transaction, no consistency gaps.

vs. Agent databases

HelixDB and HydraDB internalise a frankenstack. SurrealDB provides true multi-model unification with ACID across all data types.

GET STARTED

Start building with SurrealDB

The only database where a graph edge is a full document, a vector embedding lives in the same ACID transaction, and context is atomic.