FalkorDB

FalkorDB, Inc. makes FalkorDB, a graph database aimed at AI builders. It is open source, runs on Redis, supports OpenCypher queries, and is available as a managed cloud service or self-hosted deployment.

What It Is

FalkorDB is a graph database that stores connected data and is queried with OpenCypher. The company page describes it as a queryable property graph database that uses sparse matrices for adjacency representation and linear algebra for graph queries. It supports multi-tenant use, with thousands of isolated graphs in a single deployment so each customer, agent or session can have its own graph.

Workflows for AI applications

The site positions FalkorDB as context and memory for LLMs and agents. GraphRAG grounds LLM answers in a knowledge graph rather than vector similarity alone. Agents can query a shared model of their domain, and agentic memory keeps relationships across sessions. The homepage lists integrations with LangChain, LlamaIndex, Graphiti, Mem0 and MCP, plus a GraphRAG-SDK.

Built on the graph

QueryWeaver is a Text2SQL tool that maps a database schema as a graph so a model understands table relationships before writing SQL. Other documented graph workloads include fraud detection, cybersecurity access paths, supply chain dependency tracing, and real-time recommendations.

Setup path

A single Docker command starts FalkorDB locally, with the browser available on port 3000 for querying and visualizing graphs. Client libraries cover Python, Node, Java, Go and Rust. A managed FalkorDB Cloud and a self-hosted option with an enterprise console are also offered.

Performance claims

FalkorDB says its open benchmark against Neo4j shows 496x faster P99 latency, 6.3x leaner memory and 11.4x more throughput. The benchmark dataset, queries and code are described as public.