How does an agent create and traverse graph relationships here?

relate creates a graph edge between two records, and query runs arbitrary SurrealQL with parameter binding to traverse those edges and answer multi-hop questions. select, create, and update manage the nodes the edges connect.

Why pick SurrealDB over Neo4j for a knowledge graph?

Pick it when your data is part-graph and part-relational and you want one multi-model store. Neo4j, the top pick, is a dedicated graph database with richer traversal for graph-heavy work; SurrealDB fits when graphs sit alongside documents and tables.

SurrealDB for knowledge graphs

Pick 3 of 3 for knowledge graphsOfficialSurrealDB

For knowledge graphs, SurrealDB is our third pick of three, and it fits a particular case rather than the pure-graph one: a multi-model store that handles graph relations alongside documents and tables. Its official server lets an agent query and mutate an instance with SurrealQL and manage SurrealDB Cloud, so when your data is part-graph, part-relational, one store serves both.

Neo4j leads here as the dedicated graph database, and Memory is the lightweight entity-relation memory. SurrealDB earns the spot when you want graph edges without committing to a graph-only engine, and the relate tool plus SurrealQL give an agent a real path to walk relationships.

How SurrealDB fits

The tool that matters most for graph work is relate, which creates a graph edge between two records, paired with query, which runs arbitrary SurrealQL with parameter binding so an agent can traverse and answer multi-hop questions. select reads records from a table or by id, and create, insert, upsert, update, and delete cover building and changing the nodes. connect_endpoint and disconnect_endpoint manage the connection at runtime, while use_namespace and use_database set the active scope the queries run against.

The honest tradeoff: SurrealDB does graphs as one of several models, so for graph-heavy workloads a dedicated engine often has richer traversal and query ergonomics. Neo4j, the top pick, is built purely for graphs and its query language is shaped around them; Memory is the simple entity-relation store for lightweight agent recall rather than a full database. Choose SurrealDB when the data genuinely mixes graphs with documents and tables and you want one store the agent queries with SurrealQL, using relate to build edges and query to walk them.

Tools you would use

Tool	What it does
query	Runs an arbitrary SurrealQL statement against the connected instance, with parameter binding.
select	Selects records from a table or by record ID.
insert	Inserts one or more records into a table.
create	Creates a single record with the given content.
upsert	Inserts a record or updates it if it already exists.
update	Updates existing records with new content.
delete	Deletes records from a table or by record ID.
relate	Creates a graph edge relating two records.
connect_endpoint	Connects to a SurrealDB endpoint at runtime.
disconnect_endpoint	Disconnects from the current SurrealDB endpoint.

Full SurrealDB setup and config →

FAQ

How does an agent create and traverse graph relationships here?: relate creates a graph edge between two records, and query runs arbitrary SurrealQL with parameter binding to traverse those edges and answer multi-hop questions. select, create, and update manage the nodes the edges connect.
Why pick SurrealDB over Neo4j for a knowledge graph?: Pick it when your data is part-graph and part-relational and you want one multi-model store. Neo4j, the top pick, is a dedicated graph database with richer traversal for graph-heavy work; SurrealDB fits when graphs sit alongside documents and tables.