About SurrealDB
Overview
SurrealDB is a multi-model database and an Agent Memory layer for AI agents. It stores graphs, vectors, documents, time-series and relational data in one engine, inside a single ACID transaction. Developers query all of these data models in one language, SurrealQL.
Agent Memory connects and retrieves context wherever your data lives, and every fact carries its source. Teams can run SurrealDB self-hosted, embed it inside an application, or use SurrealDB Cloud, which manages clusters in the provider and region they choose.
Key Benefits
- Retrieves more relevant context in fewer calls, which can improve accuracy and reduce latency.
- Runs search, graph traversal, filtering and permissions in one retrieval query.
- Keeps one copy of the data, so separate systems need no synchronization.
- Defines permissions at the role, record and field level for every user, service and agent.
- Attaches a source to every fact that Agent Memory stores.
- Includes authentication, row-level permissions and live queries in the engine.
How It Works
Create a free account to start building on the database, or join the Agent Memory waitlist for early access. Connect your application data and agents to one engine, then query documents, graphs, vectors and relations together in SurrealQL. Run the engine on your own infrastructure, embed it in an application, or deploy SurrealDB Cloud in the region you choose.
Use Cases
- AI agent developers build persistent, queryable memory that tracks relationships, meaning and recency.
- Retrieval engineers build GraphRAG pipelines that combine graph traversal with vector search.
- Data modelers build knowledge graphs that query structure and meaning together.
- Digital twin teams store assets, dependencies and telemetry in one transactional engine.
- Real-time application developers use live queries so every client sees writes as they commit.
Why Choose This Product
SurrealDB keeps a database and agent memory on one engine, so retrieval avoids separate systems for graphs, vectors and documents. Agent Memory uses the database's own relationships and permissions instead of a second store. Agent Memory remains in early access, so teams join a waitlist first, while Samsung, Nvidia, Verizon, Tencent and Walmart run the database in production.
SurrealDB Pros & Cons
- Free first instance with 1GB storage on the Start tier
- Graphs, vectors, documents and relations in one ACID transaction
- Engine free to use, modify and self-host, including in production
- Built-in row-level permissions, authentication and live queries
- Production use at Samsung, Nvidia, Verizon, Tencent and Walmart
- Agent Memory is in early access and requires joining a waitlist
- Multi-model queries are marked coming soon on the Start tier
- Bring-your-own encryption keys are not included on the Start tier
Key Features
Multi-model data
All plansQueries documents, graphs, vectors, time-series and relational data in one engine and one language, SurrealQL.
ACID transactions
All plansRuns transactions across multiple rows and multiple tables with atomicity, consistency, isolation and durability guarantees.
Vector and full-text search
All plansRanks results by vector similarity and matches full text in a single statement against one copy of the data.
Agent Memory
Turns conversations, documents and business events into a knowledge graph where each fact keeps its provenance.
Row and field permissions
All plansDefines access at the table, document and field level, plus role-based access control.
Live queries
All plansSends live query notifications so every client sees a write as it commits.
Horizontal scaling
Scale+Adds compute and storage nodes independently to handle higher concurrency and multi-terabyte datasets.
Managed cloud
Runs fully managed SurrealDB clusters in the cloud provider and region you choose, scaled on demand.
Embedded and self-hosted
Runs SurrealDB embedded in Python, JavaScript, WebAssembly, mobile, edge or browser environments, or on your own infrastructure.
MCP access
Lets any model that speaks MCP reach the database and agent memory directly.
SurrealDB Pricing
View full pricing →- 1GB storage free, forever
- Scale vertically to terabytes
- Automatic burst scaling
- Daily automated backups
- Cloud RBAC and ABAC
- Further instances from $0.021/hr, billed by the hour
- Highly-available and fault-tolerant
- Multiple availability-zone deployment
- Scale horizontally towards petabytes
- Multi-region disaster recovery
- Object-storage backed data
- Self-hosted, on your infrastructure
- Clustered fault-tolerant deployments
- Audit logging and compliance
- FIPS-compliant cryptography
- Enterprise support and custom SLAs
| Feature | Start Free usage-based | Scale $0.192 usage-based | Enterprise Custom |
|---|---|---|---|
| Multi-model data | |||
| ACID transactions | |||
| Vector and full-text search | |||
| Row and field permissions | |||
| Live queries | |||
| Horizontal scaling |
Pricing extracted from the product website and may change. Check the source for current details.
Frequently asked questions about SurrealDB
What is SurrealDB?
SurrealDB is a multi-model database. Relational, document, graph, vector and time-series data are native primitives you can query together in one ACID transaction, using one query language, SurrealQL. The engine includes authentication, row-level permissions and live queries.
How is Agent Memory related to the database?
Agent Memory is a separate SurrealDB product built on the database. It turns conversations, documents and business events into a knowledge graph with provenance, using the database's own relationships and permissions rather than a second store alongside it.
Can I use SurrealDB without Agent Memory?
Yes. SurrealDB, the database, is a complete product on its own, and Samsung, Verizon, Tencent and Saks Fifth Avenue run it in production. Agent Memory is an addition for teams building agents that need persistent, governed context.
Can I self-host SurrealDB?
Yes. Run it on your own infrastructure, embed it inside your application, or use SurrealDB Cloud, which manages the service in the provider and region you choose. The same queries work across all three.
How is SurrealDB licensed?
The SurrealDB engine is free to use, modify and self-host, including in production. Its BSL 1.1 licence reserves running SurrealDB as a database service for third parties that create and manage their own schemas or tables. Each version becomes Apache 2.0 no later than four years after release, and the SDKs and tools are open source under Apache 2.0 or MIT.
Can I get started without speaking to sales?
Yes for the database: create a free account and start building. Agent Memory is in early access, and joining the waitlist lets the team open the free Sandbox to you. Enterprise support, security review and procurement are available when you need them.