Announcement

HydraDB is open source

View on GitHub

Announcement

HydraDB is open source

View on GitHub

Announcement

HydraDB is open source

View on GitHub

Star 13,008Sign up
13k
Star 13,008Sign up
13k
Star 13,008Sign up
13k

Knowledge graphs for AI agents

Knowledge graphs for AI agents

Knowledge graphs for AI agents

Turn scattered company knowledge into the right context for every agent

Start building

Start building

Book a demo

Book a demo

Trusted by teams building world-class agents

Trusted by teams building world-class agents

  • smallest.ai
  • composio
  • clinsight
  • Distyl
  • Genpact
  • Pocket
  • toptal
  • hsv.digital
  • gio
  • confida
  • klubi

Ask more of your data

Questions that reach across subjects, or ask several things about the same one

Vector databases

SQL databases

Similarity search

HydraDB knowledge graphs

Chunks with no who, why or when

Who owns it, what it touches and what that means

Old and new versions, side by side

What changed, when and why

More documents, more noise

Only what this question needs

Answers with no source

Every answer traced to its source and the decision behind it

Vector databases

SQL databases

Similarity search

HydraDB knowledge graphs

Chunks with no who, why or when

Who owns it, what it touches and what that means

Old and new versions, side by side

What changed, when and why

More documents, more noise

Only what this question needs

Answers with no source

Every answer traced to its source and the decision behind it

Vector databases

SQL databases

Similarity search

HydraDB knowledge graphs

Chunks with no who, why or when

Who owns it, what it touches and what that means

Old and new versions, side by side

What changed, when and why

More documents, more noise

Only what this question needs

Answers with no source

Every answer traced to its source and the decision behind it

Graphs keep AI accurate. HydraDB makes graphs practical

01

Automated graph creation

The ontology engine turns your scattered data into a knowledge graph of entities, relationships, sources and time

02

Built on object storage

Compute scales with traffic, not data, with no database servers holding your graph hostage. Up to 100× lower costs

03

Built in Rust

Native Cypher traversals, with recall under 200ms as your graph grows. Multi-tenant isolation, auto-scalable

04

Agent-first primitives

Agents retrieve context through MCP, CLI, SDK or your existing Neo4j drivers. Every run makes the next one better

Orchestration around the graph database

User

Request understanding

routing · entity extraction · query rewrite · safety flags

Retrieval orchestrator

Cypher reads / writes

Vectorstore (plugin)

semantic + bm25 + rerank

Connectors (plugin)

100+ sources: workspace, email, CRM

DB filters (plugin)

SQL / NoSQL

Orchestration around the graph database

User

Request understanding

routing · entity extraction · query rewrite · safety flags

Retrieval orchestrator

Cypher reads / writes

Vectorstore (plugin)

semantic + bm25 + rerank

Connectors (plugin)

100+ sources: workspace, email, CRM

DB filters (plugin)

SQL / NoSQL

The graph database core

Extremely fast, multi-tenant, and built on object storage

NAMESPACE multi-tenant isolation, auto-scalable

Writer node

Cypher writes → WAL + value log (S3 + disk cache)

Indexer node

reads WAL entries → builds index (GraphBLAS)

Reader node

Cypher queries → index + pending WAL = strongly consistent

Unified storage

S3 · object-storage native

namespace/data/wal

namespace/data/value

namespace/index

namespace/manifest.json

Tiered · Context flows across tiers on demand

Hot

in-memory cache

→

Warm

NVMe SSD

→

Cold

object storage

The graph database core

Extremely fast, multi-tenant, and built on object storage

NAMESPACE
multi-tenant isolation, auto-scalable

Writer node

Cypher writes → WAL + value log (S3 + disk cache)

Indexer node

reads WAL entries → builds index (GraphBLAS)

Reader node

Cypher queries → index + pending WAL = strongly consistent

Unified storage

S3 · object-storage native

namespace/data/wal

namespace/data/value

namespace/index

namespace/manifest.json

Tiered · Context flows across tiers on demand

Hot

in-memory cache

↓

Warm

NVMe SSD

↓

Cold

object storage

To power next-gen AI

Every agent gets context that’s connected, current, relevant and traceable

Which services depend on Atlas, who decided to move it off Postgres, when, and in which review, so we know what breaks if we roll it back?

Agents become SOTA when graphs give context

We tested retrieval with HydraDB across various industries and apps for enterprise search, code, memory and finance

Explore benchmarks
Explore benchmarks
90.79%LONGMEMEVAL-S82%BEAM 1M90.73%CODEWIKI-BENCH88.7%ENTERPRISE RAG BENCH91.4%FINANCEBENCH
90.79%LONGMEMEVAL-S82%BEAM 1M90.73%CODEWIKI-BENCH88.7%ENTERPRISE RAG BENCH91.4%FINANCEBENCH

Get started in 60 seconds

Connect your sources, build the graph and query it with openCypher

Start on Cloud

Start on Cloud

Read the quickstart ↗

Read the quickstart ↗

Step 1

Connect

Docs, chat, email, CRM and code, through 100+ connectors

  • chats
  • emails
  • docs
  • CRM records
  • memory
  • 100+ apps

Step 2

Ingest

Entities and relationships are extracted on ingest. No schema to write

hydradb ingest ./docs/*.md

Step 3

Query

openCypher over Bolt, with the Neo4j drivers you already use

1MATCH (p:Project {name: "Atlas"})2 -[:DEPENDS_ON]->(d)3RETURN d.name

Developers are building with HydraDB
12K+ GitHub stars

Explore GitHub
Explore GitHub

pulling off sub-200ms recall on a full relational graph is an INSANE engineering flex.
props to @contextkingceo and @hydra_db - If these latency numbers hold up, this launch is pretty massive

Charly Wargnier

@DataChaz

I built my own mini @hydra_db
Spent the last few weeks rebuilding an AI memory system from scratch, this time based on the Hydra DB paper.

Karthiknadar1204

@uchiha_katsuki_

Sometimes you see a new technology so disruptive that it can kill an entire industry.
Hydra DB is not just the replacement for a vector DB but will also completely change how the world thinks of databases.

Sudarshan Kamath

@kamath_sutra

Of everything agent-related I’ve seen lately, HydraDB is the only one my first reaction said is solid. Knowledge graph + log structure. You can tell at a glance it was designed by someone who understands both algorithms and computer systems.

Menci

@lcMenci · translated from Chinese

building Errata: an append only belief-graph on @hydra_db for AI Agents memory
- corrections don’t overwrite old facts
- they become new edges that supersede them
- so agents can track what they believed, what changed, and when

Aman Jha

@WhyParabola

When a package is compromised, you have minutes.
#Radius answers two critical questions: what’s exposed and what’s the smallest fix?
It does this through real graph traversal over @hydra_db, not a lookup or a scan.
Tested against the real May 2026 TanStack attack

Jurban Ph

@jurbanwrites

@Maaztwts built OSAP to stop the constant tab-switching between Slack, GitHub, and Notion.
It uses GLM-5.1 as a reasoning layer to handle multi-step tasks across different apps, with persistent memory (HydraDB) to keep track of your specific workflow habits.

Z.ai

@Zai_org

Built TESSERA: 511,962 enterprise docs across 9 sources, one graph.
Instead of guessing identity from embeddings, it traces shared neighborhoods in @hydra_db
91% accuracy on EnterpriseRAG-Bench. Graph-native ontology.

Kinnho

@0xkiddok

I built Orange: Google for a made-up company, where your clearance decides what you can find and do. Each person can talk to the Company Brain agent with HydraDB limiting what they have access to before the AI agent ever reads it

Jason

@jasonjias

Frequently asked questions

Explore documentation
Explore documentation

HydraDB is a graph database built on object storage, so your graph can grow without tying storage to database servers. Query and indexing capacity scale independently, giving you infrastructure built for large, connected datasets and the agents that depend on them

HydraDB is for engineers building products that need rich, connected context. Context can be scattered across emails, chats, docs, transcripts, PDFs, code and memories. Graphs connect it all, so your agent stays on top of it

You can. Postgres with a graph extension is a practical starting point. As your graph and query volume grow, graph indexes and traversals can compete with your core database for memory and CPU, slowing everyday transactions. HydraDB runs the graph workload on separate, purpose-built infrastructure, so you can scale graph queries without consuming the resources your application depends on

Use a vector database when you mainly need to find similar documents or passages. Use HydraDB when your application needs to follow relationships, traces and context across multiple places: who owns a project, what depends on it, or which past decisions explain its current state. We don’t ask you to replace your vector database. Both graph search and vector search serve different purposes. In fact, HydraDB comes with native support for vector search

Yes. HydraDB can serve as the graph database behind your GraphRAG application, storing entities and relationships your retrieval pipeline can query. Keep your models and extraction pipeline, then use HydraDB to retrieve connected context for questions that require following relationships across multiple sources

HydraDB is a graph database built on object storage, so your graph can grow without tying storage to database servers. Query and indexing capacity scale independently, giving you infrastructure built for large, connected datasets and the agents that depend on them

HydraDB is for engineers building products that need rich, connected context. Context can be scattered across emails, chats, docs, transcripts, PDFs, code and memories. Graphs connect it all, so your agent stays on top of it

You can. Postgres with a graph extension is a practical starting point. As your graph and query volume grow, graph indexes and traversals can compete with your core database for memory and CPU, slowing everyday transactions. HydraDB runs the graph workload on separate, purpose-built infrastructure, so you can scale graph queries without consuming the resources your application depends on

Use a vector database when you mainly need to find similar documents or passages. Use HydraDB when your application needs to follow relationships, traces and context across multiple places: who owns a project, what depends on it, or which past decisions explain its current state. We don’t ask you to replace your vector database. Both graph search and vector search serve different purposes. In fact, HydraDB comes with native support for vector search

Yes. HydraDB can serve as the graph database behind your GraphRAG application, storing entities and relationships your retrieval pipeline can query. Keep your models and extraction pipeline, then use HydraDB to retrieve connected context for questions that require following relationships across multiple sources

HydraDB is a graph database built on object storage, so your graph can grow without tying storage to database servers. Query and indexing capacity scale independently, giving you infrastructure built for large, connected datasets and the agents that depend on them

HydraDB is for engineers building products that need rich, connected context. Context can be scattered across emails, chats, docs, transcripts, PDFs, code and memories. Graphs connect it all, so your agent stays on top of it

You can. Postgres with a graph extension is a practical starting point. As your graph and query volume grow, graph indexes and traversals can compete with your core database for memory and CPU, slowing everyday transactions. HydraDB runs the graph workload on separate, purpose-built infrastructure, so you can scale graph queries without consuming the resources your application depends on

Use a vector database when you mainly need to find similar documents or passages. Use HydraDB when your application needs to follow relationships, traces and context across multiple places: who owns a project, what depends on it, or which past decisions explain its current state. We don’t ask you to replace your vector database. Both graph search and vector search serve different purposes. In fact, HydraDB comes with native support for vector search

Yes. HydraDB can serve as the graph database behind your GraphRAG application, storing entities and relationships your retrieval pipeline can query. Keep your models and extraction pipeline, then use HydraDB to retrieve connected context for questions that require following relationships across multiple sources

HydraDB is a graph database built on object storage, so your graph can grow without tying storage to database servers. Query and indexing capacity scale independently, giving you infrastructure built for large, connected datasets and the agents that depend on them

HydraDB is for engineers building products that need rich, connected context. Context can be scattered across emails, chats, docs, transcripts, PDFs, code and memories. Graphs connect it all, so your agent stays on top of it

You can. Postgres with a graph extension is a practical starting point. As your graph and query volume grow, graph indexes and traversals can compete with your core database for memory and CPU, slowing everyday transactions. HydraDB runs the graph workload on separate, purpose-built infrastructure, so you can scale graph queries without consuming the resources your application depends on

Use a vector database when you mainly need to find similar documents or passages. Use HydraDB when your application needs to follow relationships, traces and context across multiple places: who owns a project, what depends on it, or which past decisions explain its current state. We don’t ask you to replace your vector database. Both graph search and vector search serve different purposes. In fact, HydraDB comes with native support for vector search

Yes. HydraDB can serve as the graph database behind your GraphRAG application, storing entities and relationships your retrieval pipeline can query. Keep your models and extraction pipeline, then use HydraDB to retrieve connected context for questions that require following relationships across multiple sources

HydraDB is a graph database built on object storage, so your graph can grow without tying storage to database servers. Query and indexing capacity scale independently, giving you infrastructure built for large, connected datasets and the agents that depend on them

HydraDB is for engineers building products that need rich, connected context. Context can be scattered across emails, chats, docs, transcripts, PDFs, code and memories. Graphs connect it all, so your agent stays on top of it

You can. Postgres with a graph extension is a practical starting point. As your graph and query volume grow, graph indexes and traversals can compete with your core database for memory and CPU, slowing everyday transactions. HydraDB runs the graph workload on separate, purpose-built infrastructure, so you can scale graph queries without consuming the resources your application depends on

Use a vector database when you mainly need to find similar documents or passages. Use HydraDB when your application needs to follow relationships, traces and context across multiple places: who owns a project, what depends on it, or which past decisions explain its current state. We don’t ask you to replace your vector database. Both graph search and vector search serve different purposes. In fact, HydraDB comes with native support for vector search

Yes. HydraDB can serve as the graph database behind your GraphRAG application, storing entities and relationships your retrieval pipeline can query. Keep your models and extraction pipeline, then use HydraDB to retrieve connected context for questions that require following relationships across multiple sources

Give your agents the
context they’re missing

Self-host it free, or let us run it for you on HydraDB Cloud

Start on Cloud

Start on Cloud

Self-host from GitHub ↗

Self-host from GitHub ↗

GitHub stars

13k

Retrievals per month

1M

Improvement in agent recall

4x

Trusted by

2k devs

GitHub stars

13k

Retrievals per month

1M

Improvement in agent recall

4x

Trusted by

2k devs