HelixDB

HelixDB is an open-source graph-vector database built from scratch in Rust, designed to consolidate the storage needs of AI applications into a single platform. It is developed by Helix DB, Inc., backed by Y Combinator and Nvidia, and is available both as a self-hosted open-source project (Apache 2.0) and as a managed cloud service. The repository has accumulated nearly 5,000 GitHub stars since its creation in November 2024.

What It Is

HelixDB is an OLTP database that combines graph, vector, key-value, document, and relational data models in one system. The core thesis is that AI applications — particularly those requiring RAG pipelines, agent memory, and knowledge graphs — should not need to stitch together separate application databases, vector stores, and graph databases. HelixDB gives agents and applications federated access to all of these through a single query interface. The primary data model is graph + vector, but the system also supports KV, document, and relational data.

Architecture and Deployment Model

HelixDB runs in two modes: local (self-hosted via CLI) and Helix Cloud (managed, object-storage-backed).

• Local/self-hosted: Installed via a one-line curl script; runs as a background container on port 6969. Supports in-memory (default) and disk-persistent modes.
• Helix Cloud: Built on object storage to handle any volume of graph and vector data. Uses a single-writer, auto-scaling reader architecture behind a gateway layer. The gateway buffers requests so DB node failures do not surface as errors. High availability requires at least 3 gateways and 3 DB nodes; smaller clusters run in Dev mode.
• ACID transactions: Full ACID support is included in the cloud deployment.

Developer Workflow and SDKs

Queries are authored using a Rust or TypeScript DSL and sent as dynamic requests to a running instance via POST /v1/query — no build or deploy step is required. Both SDKs produce the same JSON AST.

• Rust SDK: Published as helix-db on crates.io; queries are defined as #[register] functions.
• TypeScript SDK: Published as @helix-db/helix-db on npm; supports Node.js 20+.
• helix chef: An interactive bootstrapper that scaffolds a project, starts a local instance, seeds example data, and can hand off to a coding agent (Claude Code, Codex, or OpenCode) to build a working application from a one-line description.
• HQL: A native query language documented separately from the DSL.
• MCP integration: The CLI installs HelixDB query skills and a docs MCP server as part of the helix chef setup.

Update: v3.0.5

The latest release is v3.0.5, published on June 5, 2026. The repository was last pushed to on June 10, 2026, indicating active development. The project was created in November 2024 and has grown to nearly 5,000 stars and 264 forks. The GitHub topics include rag, vector-database, graph-database, and ai, reflecting the product's positioning at the intersection of AI infrastructure and database tooling.

Open-Source Lineage and Backing

The core HelixDB engine is released under the Apache License 2.0, making it freely usable, modifiable, and distributable. The project is backed by Y Combinator and Nvidia, and the company is headquartered across London and San Francisco. A SOC 2 trust center is available at trust.helix-db.com. Commercial support is available through Helix Cloud, with direct founder contact offered for enterprise inquiries.