Description
HelixDB is an open-source graph and vector database built for AI memory. Its data model is a property graph, and vector search and full-text (BM25) search are indexes over the properties of nodes and edges within the same engine, so there is no separate graph store, vector store or search cluster to keep in sync. The engine is written in Rust and runs on object storage. The site describes three main uses: agent memory, where memories are stored as nodes and edges and recalled by traversing to what an agent is allowed to see and then searching inside that set; a company knowledge graph that connects code, documents, people and projects, so answers link to the owner and source; and retrieval for AI applications, grounding answers in records each user is allowed to see. Permissions are enforced by prefiltered search: a graph traversal or indexed tenant property defines the candidate set, and ranking runs only inside it in the same request. HelixDB does not generate embeddings or rerank results; your application computes embeddings and combines the vector and text results. Official SDKs exist for Rust, TypeScript, Python and Go, and there is no text query language, with queries built in code or sent as JSON to an endpoint. The engine can be embedded in a process, self-hosted with Docker, deployed from cloud marketplaces, run bring-your-own-cloud, or used as a managed service. The site notes that the earlier HelixQL language is archived, and it is backed by Y Combinator.…
Preview
Highlights
- Graph and vector search in one engine
- Full-text BM25 search
- Prefiltered permission-aware search
- Official SDKs for Rust, TypeScript, Python and Go
- Embedded, Docker, BYOC or managed
- Built for agent memory
Ratings & Reviews
No ratings yet
Be the first to rate HelixDB.
People also visit
HelixDB FAQ
What is HelixDB?
An open-source graph-vector database for AI memory, with vector and full-text search in one Rust engine.
Does it generate embeddings?
No. The site says your application computes embeddings and combines vector and text results, while HelixDB stores and searches them.
How are permissions enforced?
Through prefiltered search, where a graph traversal or tenant index defines the candidate set before ranking.
What languages can I use?
Official SDKs for Rust, TypeScript, Python and Go. There is no text query language to learn.
Can I self-host it?
Yes. The engine can be embedded, run in Docker, or deployed bring-your-own-cloud, and a managed option is also offered.
Is the old query language still used?
The site says HelixQL is archived.
What are the best HelixDB alternatives?
Popular alternatives to HelixDB include Memgraph, RexaDB and TerminusDB. See all HelixDB alternatives




