v0.6.0 — Agentic RAG is here

The Knowledge Layer
for AI Agents

A hypergraph database built for multi-hop reasoning. Connect entities through semantic relationships and let your agents traverse knowledge like never before.

Get Started Free Read the Docs

agentic_rag.py

from hyperx import HyperX
from hyperx.agents import create_tools

# Initialize with your knowledge graph
client = HyperX(api_key="hx_sk_...")

# Create agent tools with quality signals
tools = create_tools(client, level="explore")

# Execute with self-correction
result = tools.execute(
    "hyperx_search",
    query="React state management"
)

if result.quality.should_retrieve_more:
    # Quality signals guide agent behavior
    print(result.quality.suggested_refinements)
            

Core Capabilities

Beyond Traditional Databases

HyperX rethinks how AI systems should store and retrieve knowledge with hypergraph semantics.

⬡

Hyperedges

Unlike edges in traditional graphs, hyperedges connect any number of entities with rich semantic roles. Model complex relationships like "Author wrote Book published by Publisher in 2024."

↗

Multi-Hop Paths

Traverse knowledge graphs with configurable depth. Find connections between concepts through semantic pathways with role-aware filtering at each hop.

🔍

Hybrid Search

Combine vector similarity with keyword matching and graph structure for precise retrieval.

📊

Bi-Temporal

Track both when facts were true and when they were recorded. Essential for audit trails.

🧠

Native Embeddings

Built-in vector storage without external dependencies. 1536-dimension vectors supported.

v0.6.0 Feature

Agentic RAG

Purpose-built tools for AI agents with quality signals that enable self-correction and multi-step reasoning.

Quality-Aware Retrieval

Every tool response includes quality signals that help agents decide when to retrieve more data, refine queries, or proceed with reasoning.

Confidence scoring for result reliability
Coverage metrics across entity types
Diversity signals to avoid echo chambers
Automatic query refinement suggestions
Self-correction triggers for agents

8 Tools across 3 Access Levels

SearchTool PathsTool LookupTool ExplorerTool ExplainTool RelationshipsTool EntityCrudTool HyperedgeCrudTool

read → Search, Paths, Lookup
explore → + Explorer, Explain, Relationships
full → + EntityCrud, HyperedgeCrud

Framework Support

Works With Your Stack

Native integrations for popular LLM frameworks. Zero boilerplate, full type safety.

🦜

LangChain

HyperXToolkit for LangGraph agents

from hyperx.agents import HyperXToolkit
toolkit = HyperXToolkit(client, level="explore")
tools = toolkit.get_tools()

🦙

LlamaIndex

HyperXToolSpec for agent workflows

from hyperx.agents import HyperXToolSpec
spec = HyperXToolSpec(client, level="explore")
tools = spec.to_tool_list()

🤖

OpenAI Functions

Export schemas for function calling

tools = create_tools(client)
schemas = tools.schemas # OpenAI format
tools.execute("hyperx_search", query="...")

Simple Pricing

Start Free, Scale As You Grow

No credit card required. Upgrade when you need more power.

Free

$0 / forever

Perfect for prototypes and side projects

1,000 entities
5,000 hyperedges
Full SDK access
Agentic RAG tools
Community support

Get Started

Pro

$29 / month

For production applications

Unlimited entities
Unlimited hyperedges
Priority support
Custom embeddings
Team collaboration
SLA guarantee

Start Free Trial

The Knowledge Layer for AI Agents