SurrealDB review

SurrealDB is a database built to hold several kinds of application data in one engine: relational, document, graph, vector and time-series. Instead of choosing separate systems for each model, teams can query across them with SurrealQL, which covers SQL-style queries as well as graphs, vectors, full text, temporal data and key-value data. The product also says it supports PostgreSQL and MySQL clients.

The pitch extends beyond storage. SurrealDB lists ACID transactions, role-, record- and field-level access controls, and live queries that notify clients as writes commit. Its Agent Memory feature stores and retrieves agent context alongside its source and time. You can self-host or use SurrealDB Cloud, with serverless Earth plans and always-on Cloud Instances. Our verdict: it is worth evaluating when consolidating data models is a real requirement, but breadth alone is not a reason to replace a database you already understand.

The appeal is fewer moving parts when an application genuinely needs different data shapes. A team can keep relational records, nested documents, graph links and vectors in one database rather than maintaining separate engines for each. SurrealQL is the common query interface, and the product lists ACID transactions for combining operations in a transactional query. Live queries may also suit applications that need clients to receive updates as writes commit.

The Cloud entry point is relatively clear. According to the pricing page, Earth Free costs $0 per month and includes 3 million requests, 10 GB of storage and 100 database hours monthly. It requires no credit card; at its limits, the database pauses while data is kept. Earth Pro costs $15 per month and includes 10 million requests, 25 GB of storage, 1,000 database hours, 500 million live update messages and 10,000 connected clients.

For agent applications, Agent Memory’s ability to store context with its source and time is a concrete feature, not just a general AI label. That could help teams track where retrieved context came from.

Combining data models also concentrates risk: if SurrealDB does not fit a workload, the convenience of one engine may not make up for it. Reddit discussions include old, user-reported performance concerns, including one evaluation of version 1.2.0 that described slower results than SQLite for its particular workload. That is not a current benchmark, and it does not establish how SurrealDB performs today, but it is a reason to test your own queries before committing.

The free plan is tightly capped: 3 million requests, 10 GB and 100 database hours per month, after which it pauses. Earth Pro adds capacity, but overages cost $0.65 per million requests, $0.25 per GB of storage and $0.05 per active database hour. Watch those meters if usage is hard to predict.

Cloud Instances start at $0.021 per node hour, according to the pricing page; total cost depends on usage, and no fixed monthly price is published. Discounted annual Cloud and support plans are mentioned, but prices are not published. Enterprise pricing is custom, and listed optional add-ons—including SSO and advanced support—have no published prices.

SurrealDB’s Earth Free plan costs $0 and includes 3M requests, 10 GB of storage and 100 database hours per month. It pauses at those limits, but keeps your data. Earth Pro is $15 per month for 10M requests, 25 GB and 1,000 database hours, plus 500M live-update messages and 10,000 connected clients. That’s a substantial allowance increase for a fixed monthly price, but it is not an uncapped plan.

Earth Pro overages cost $0.65 per million requests, $0.25 per GB of storage and $0.05 per active database hour. Check expected usage before moving production workloads: those charges can make the monthly bill less predictable. Cloud Instance pricing starts at $0.021 per node hour; total cost depends on usage, and the pricing page does not publish a fixed monthly price. Enterprise pricing is custom. The page mentions discounted annual plans for Cloud and support, but does not list annual prices. Optional add-ons—including SSO and Cloud private networking—are listed without prices.

Treat a move to SurrealDB as a data-model and application change, not just a database swap. It combines relational, document, graph, vector and time-series data, but applications using SurrealQL features such as graph queries, live queries or record- and field-level permissions may need changes to use those capabilities. SurrealDB also accepts Postgres and MySQL clients; that does not establish that every query or application will transfer unchanged.

Before committing, map schemas, queries, access rules and agent-memory data, then confirm how to export and restore them. Export formats and a migration path are not specified on the vendor's site material, so we can’t promise portability. The same caution applies when leaving: check that your target database can represent the data and behavior you rely on. SurrealDB’s native models and features may create switching work if the replacement lacks equivalents. The published material includes no Reddit reports about why users leave, so we can’t attribute migration complaints or exit reasons to Reddit.