What it does HelixDB is an open-source database built from scratch in Rust that combines graph and vector storage with KV, document, and relational support. It targets AI applications—specifically RAG, agents, and “company brains”—that typically need separate vector DBs, graph DBs, and application databases. The project ships a CLI (helix) for local development and a managed cloud option with ACID transactions and auto-scaling readers.

The interesting bit The query model is genuinely unusual: you write queries in a Rust or TypeScript DSL, register them with attributes or plain functions, and the SDK serializes them to a JSON AST that gets POSTed to POST /v1/query. No compile-deploy loop for local work. The helix chef bootstrapper will even scaffold a full app from a one-liner and hand off to Claude Code or Codex if it detects one.

Key highlights

• Built-in MCP tooling so agents can traverse the graph directly instead of generating human-readable queries
• Native embedding via an Embed function—no pre-processing pipeline required
• LMDB-backed storage engine; the README claims “ultra-low latency” but offers no numbers
• AGPL license with commercial cloud offering (single writer, multiple readers, object storage backend)
• helix run dev defaults to in-memory; --disk persists across restarts

Caveats

• The “ultra-low latency” claim is unsupported by benchmarks in the README
• Default local mode is in-memory, which is a footgun if you miss the --disk flag
• Cloud and local instances use different deploy paths (helix sync vs. helix push), adding friction if you outgrow your laptop

Verdict Worth a look if you’re tired of gluing Neo4j, Pinecone, and Postgres together for every AI prototype. Skip it if you need mature operational tooling, detailed performance data, or a simple hosted option without talking to sales.