SurrealDB 2.0 and the pivot from MongoDB

Consolidation of fragmented stacks

Tencent replaced nine backend tools with SurrealDB in 2024 to power its infrastructure monitoring platform. This change reduced complexity and lowered the total cost of ownership. Aspire Competitions also consolidated five separate tools into one in 2025 to increase development cycles. Many organizations manage 3-4 databases on average, which adds extra complexity. A March 2025 survey by Matillion found that 64% of organizations report their data teams spend more than 50% of their time on repetitive or manual tasks. An early 2025 report by IDC found that in 2024, only about 16% of developers’ time was spent on actual application development. Most engineering effort goes into integrating multiple databases, writing glue code, and managing schema migrations. SurrealDB provides a single engine for documents, relationships, semantics, embeddings, and event-driven logic to stop this fragmentation. The company has reached $38 million in total funding. The project reached over 32.9k stars on GitHub as of August 2026.

Comparison of MongoDB and SurrealDB

MongoDB is a mature document database that has been in production since 2009. It stores BSON documents in collections and uses JSON-based query APIs and aggregation pipelines. To handle real-time updates, the platform requires change streams and a replica set. Because it scales via sharding, MongoDB handles petabytes of data. Its ecosystem includes Compass, Atlas, and Mongoose. SurrealDB is a multi-model database built in Rust that combines document, graph, relational, time-series, geospatial, and key-value data. SurrealDB includes built-in live queries that push updates to clients via WebSockets.

Because SurrealDB supports multiple paradigms, highly specialized tools may outperform it in extreme, narrowly focused workloads like large-scale OLAP stores or ultra-optimized graph traversal for drug discovery. Engineers who use MongoDB often spend significant time writing glue code, managing schema migrations across many different services, and maintaining separate security layers for each data store in their fragmented and complex software stacks. SurrealDB uses SurrealQL, which blends SQL syntax with graph semantics. This allows users to start schemaless and move to schema enforcement as models mature. It also allows users to connect records through directed graph edge connections. This eliminates the N+1 query problem because inter-document record links allow for performant related queries without the use of JOINs. SurrealDB also provides row-level permissions and auth scopes that tie to JWTs or sessions. This removes the need for custom authentication glue in backend services.

Deployment flexibility and AI implementation

SurrealDB works as a single Rust binary. It runs in-memory, on the edge, or embedded in a browser via WebAssembly. This makes it suitable for IoT and edge use cases. Developers can use it on mobile or in offline-first applications. Verizon used the platform in 2024 to build a generative AI assistant for its 10,000 field technicians. The technicians use the tool to access technical documentation and outage updates. This is possible because SurrealDB supports vectors, text, and graph relationships for AI agent context. It enables scenarios like Retrieval Augmentation Generation and knowledge graphs. The system also handles multimodal data like images and audio. Users can install it via Homebrew on macOS or use Docker for container orchestration.

AI applications rely on vectors, text, and graph relationships for agent context and memory. SurrealDB supports these models natively to enable Retrieval Augmentation Generation and knowledge graphs for AI. It also provides support for multimodal data like images and audio. This allows developers to build recommendation engines and fraud detection systems without extra tools.

Building SurrealDB crates causes long compile times for some developers. One developer reported that a full build of a workspace took over 5 minutes on a MacBook M3. Another developer found that a build ran for 50 minutes on a DigitalOcean droplet before it crashed because the system lacked memory. This makes it hard to build on a Raspberry Pi.

The pricing for SurrealDB starts with a free plan providing 1GB of storage and 1GB of memory. The Start plan costs $0.021 per hour for continuous use. If your workload requires extreme horizontal sharding at petabyte scale, MongoDB is the more proven choice. You should evaluate whether your data model is naturally graph-like before switching. Can a single multi-model engine truly replace a dozen specialized database tools for every enterprise use case?