Hello Product Hunt community 👋

We’re excited to open source HydraDB today.

What is HydraDB?

Our goal is to support modern AI use cases like agent memory, company brain, and ontologies. A lot of agents today are built with vectorDBs or stitch together multiple around semantic similarity: embed something, retrieve closest chunks, and pass to models.

But similarity isn’t always relevance. Moreover, information in isolation doesn’t work well for agents. They come back with more questions about that single chunk than answers or work done.

HydraDB is a graph database, built on object storage, written in Rust, for intelligent AI that requires context extremely fast, maintains state over time, all while making it extremely cost efficient.

Context an agent needs is a notion page connected to your emails, messages, and 10 other documents so that your agent gets to choose how much context is enough to complete some work. That’s when it gets a “brain” that hold a relationship, a dependency, a sequence of events, a previous decision, or a fact that has since been superseded. It’s an intelligent graph.

Why did we build it?

As we worked on AI systems, we kept running into the same problem: models were getting better and eventually reaching a point where intelligence is no longer a bottleneck, but the context they received was still fragmented and lossy.

Enterprise data lives across Slack, Jira, Gmail, GitHub, Drive and dozens of other systems. Agents need to understand not just the individual pieces of data, but how those pieces relate to each other. The same thing shows up in agent memory, code intelligence, knowledge systems, security, and other domains.

We think graphs are going to become an increasingly important part of the context layer for AI. We built HydraDB to make your AI relationally aware, give it connected context easier to store, retrieve, and reason over.

So what are we launching today?

Today we’re open-sourcing HydraDB.

You can now run it yourself, build on top of it, inspect how it works, and tell us where it breaks.

We’re especially excited to see what people build around:

agent memory and long-term context
GraphRAG and knowledge retrieval
enterprise knowledge and ontologies
code and dependency graphs
multi-hop reasoning

Why open source?

Because we don’t think the next generation of AI infrastructure should be something developers have to take on faith. AI needs to be sovereign and something you own.

If HydraDB is going to become useful infrastructure, developers should be able to run it themselves, understand it, extend it, benchmark it, and push it into use cases we haven’t thought of.

We also know we won’t figure this out alone.

A big reason for this launch is to get HydraDB into the hands of people building real systems and learn from what they do with it.

So if you try it, we’d love to hear:

What did you build?
What worked?
What was painful?
And what would make HydraDB dramatically more useful for you?

Thanks for checking us out and special s/o to everyone who has helped us get here ❤️