memgraph/memgraph

What is Memgraph?

Memgraph is an in‑memory graph database written in C/C++. It stores nodes and edges in RAM, which lets it answer complex multi‑hop graph queries in sub‑millisecond time. While it works as a general‑purpose graph engine, the project is positioned specifically for AI‑centric workloads such as GraphRAG, AI memory stores, and agentic pipelines that need fast, combined graph‑plus‑vector or text search.

---

Key Features (as described in the README)

Area	What it does
AI & Graph Intelligence	• Built‑in vector, text, and geospatial indexes for similarity and keyword search. • MAGE library – 40+ graph algorithms (PageRank, community detection, GNN‑based link prediction, temporal graph networks, embeddings, etc.) available in C++, Python and CUDA. • Atomic GraphRAG – a single Cypher query can perform pivot search, graph expansion, ranking and prompt assembly. • LLM utility module for formatting graph‑aware context for large language models. • AI Toolkit integration with popular agent frameworks and ready‑made components for GraphRAG, AI memory and agent workflows. • Real‑time schema introspection (SHOW SCHEMA INFO).
Performance & Query Power	• Pure in‑memory C/C++ engine delivering sub‑millisecond traversals (benchmarked). • Deep‑path traversals with accumulators and path filtering. • Extensible via custom query modules written in Python, Rust, or C/C++. • Parallel query execution for high throughput. • Native loading of Parquet & JSONL from local disk, S3 or HTTP. • Streaming ingestion from Kafka, Pulsar, RedPanda with dynamic algorithms that react to changes.
Enterprise‑grade capabilities	• Raft‑based high availability with automatic failover. • Multi‑tenancy and fine‑grained role/label‑based permissions. • SSO, user impersonation, 30+ permission types. • Encryption in transit, monitoring, backup & restore.
Tooling & Ecosystem	• Memgraph Playground – a browser‑based sandbox (no install needed). • Memgraph Lab UI for visual exploration and query tuning. • Drivers for Python, C/C++ and others, plus WebSocket access. • Helm charts for Kubernetes deployments (stand‑alone or HA).

---

Who Might Use It?

• AI engineers building Retrieval‑Augmented Generation (RAG) pipelines that need to combine vector similarity with graph reasoning in a single query.
• Data scientists running real‑time graph analytics for fraud detection, network monitoring, or infrastructure analysis where latency matters.
• Developers of agentic systems that require fast context lookup, schema introspection, and seamless integration with LLMs.
• Enterprises that need a high‑availability, ACID‑compliant graph store with strong security and multi‑tenant isolation.

---

How to Get Started

1. Try the free Playground – visit https://playground.memgraph.com/ for an instant, no‑install demo.
2. Install locally – Docker, WSL, or native packages are provided for Windows, macOS and Linux (links in the README).
3. Deploy to Kubernetes – add the Helm repo (helm repo add memgraph https://memgraph.github.io/helm-charts) and install with helm install.
4. Connect – use Memgraph Lab, mgconsole, or any of the language drivers (Python, C/C++, etc.).
5. Explore AI features – check the src/mage folder for algorithm implementations and the separate ai-toolkit repo for ready‑made GraphRAG components.

---

License & Community

• Community edition – Business Source License (BSL).
• Enterprise edition – Memgraph Enterprise License (MEL).
• Active Discord, Stack Overflow tag, Twitter, and YouTube channels.
• Contribution guidelines, code of conduct, and a “best‑practices” repo are provided for developers who want to help.

---

In short: Memgraph is a high‑performance, in‑memory graph database that bundles vector/text search and a rich set of graph algorithms, making it a solid foundation for modern AI‑driven applications that need fast, connected context retrieval.