HydraDB vs HelixDB
HelixDB is an open-source graph database with vector and BM25 search, built in Rust on object storage. Your app computes the embeddings, merges the results and models history, and its Foundry service builds graphs from documents and developer tools.
When agents need the graph built and kept current, with every change versioned, HydraDB does that for you.
Based on public docs and pricing as of Nov 2026. Something out of date? Suggest a correction →
HelixDB is an open-source graph database with native vector and BM25 full-text search, built in Rust on object storage and licensed Apache 2.0. Each request runs as one ACID transaction. Queries go through typed SDK builders or JSON, with an openCypher profile and SQL on top, and Helix Cloud adds an MCP server for agents. It runs embedded, in Docker or on Helix Cloud.
It's the go-to for teams that want a permissively licensed graph and vector database they can embed or self-host, and that already run their own embedding pipeline.
The friction shows up when nobody wants to build that pipeline, or when the agent needs to know what changed.
The caveats worth knowing
Embeddings and ranking are your code. HelixDB doesn't generate embeddings or rerank. If you run both vector and keyword search, your app combines the results.
The graph is what your app writes. The database stores the nodes and edges you send it. Turning Slack threads, emails and docs into entities and relationships is extraction code you write and run, unless you add Foundry.
History is yours to model. There's no query for how a record looked before. Tracking versions means isLatest flags and edges you design.
What changes with HydraDB?
Search comes back ranked in one query. Vector search, BM25 keyword search and graph traversal run together, with metadata filters and recency weighting. There's no embedding pipeline to run.
The graph is created from your data, and keeps evolving. 100+ connectors keep Slack, Gmail, Drive, Jira and others in sync, permissions included. Entities and relationships are created as data arrives, structured to your ontology or one proposed for you, and the graph restructures when new kinds of facts appear.
Agents can find what changed, when, and why. Every change to the graph is versioned like code in Git, with when, why and where it came from.
What about Helix Foundry?
Foundry is HelixDB's hosted service for building graphs. It drafts an ontology across your documents and sources, uses Claude models to find the names and facts they hold, then loads the result into a HelixDB database. It connects to GitHub, Linear, Stripe, PostHog, Postgres, MySQL, S3 and more.
It's the go-to for teams building graphs from product and developer data, like code, issues, billing and analytics, with HelixDB underneath.
The difference shows up when the knowledge lives in conversations, email and docs.
Workplace apps connect directly. Foundry's connector list has no Slack, Gmail or Drive. HydraDB syncs those and 100+ others.
The structure keeps up with the data. Foundry's types and links change only when you generate again, and each load writes a complete new copy of the graph. HydraDB updates the graph incrementally and restructures it when new kinds of facts appear.
Technical comparison
| HelixDB | Helix Foundry | HydraDB | |
|---|---|---|---|
| What it is | Open-source graph database with vector and BM25 search | Hosted service that builds graphs into HelixDB | Graph infrastructure: storage, creation, evolution and retrieval |
| Storage | Written in Rust; data on object storage, with NVMe and memory caches; one writer node, reader nodes scale out | Writes into a HelixDB database | Written in Rust; the graph lives on object storage, with memory and NVMe as caches; compute and storage scale separately |
| Graph creation | Nodes and edges your app writes | Claude models draft an ontology and extract facts from documents and connected sources | From 100+ app connectors, file uploads and the API, structured to your ontology or a proposed one |
| Keeping it current | Your writes | Syncs daily (Free), hourly (Pro) or every 15 minutes (Max); each load writes a complete new copy; types change only when you generate again | Updates incrementally as data arrives and restructures when new kinds of facts appear |
| Retrieval | Traversal, with vector and BM25 search scoped to its results; embeddings and result fusion in your app | Queries run on the HelixDB database it loads | Vector search, BM25 keyword search and graph traversal combined in one query, with metadata filters and recency weighting |
| Query interface | Typed SDK builders or JSON; openCypher profile without shortest or variable-length paths; SQL; MCP on Helix Cloud | Hosted workspace | Natural language through the API and SDKs; openCypher on the open-source database |
| When a fact changes | New value replaces the old; versions modelled with flags | Not documented | Every version kept, Git-style, with its source |
| Deployment | Embedded, Docker, or Helix Cloud; bring your own cloud coming soon | Hosted | Managed, dedicated, your cloud, or self-hosted |
| Pricing | Cloud from $5/month (200K reads) to $200/month (10M reads); dedicated clusters billed hourly (pricing) | Free; Pro $100/month; Max $500/month | Managed from $25/month: $0.70 per 1M tokens ingested, $0.50 per GB-month storage; open source free (pricing) |
| License | Apache-2.0 | Hosted service | Elastic License 2.0 |