Memgraph

Memgraph develops an in-memory graph database written in C/C++ that stores data in the property graph model of nodes and relationships with key-value properties, and queries it with the Cypher/openCypher query language. The engine keeps working data in RAM to avoid disk I/O latency while persisting through periodic snapshots and a write-ahead log, and it supports ACID transactions with multi-version concurrency control and snapshot isolation, skip-list indexes, and Raft-based replication and coordination with automatic failover. The product is positioned around two workload families: AI context — GraphRAG retrieval, long-term AI memory (semantic, episodic and procedural) and agentic reasoning graphs — and real-time graph analytics such as fraud detection, network analysis, data lineage, knowledge graphs, identity and access management and supply chain modelling.

The platform combines built-in vector, text and geospatial indexes with graph traversal so hybrid retrieval can run inside a single Cypher query, and ships the MAGE library of 40+ graph algorithms in C++, Python and CUDA including PageRank, community detection, GNN-based link prediction, temporal graph networks and embeddings. Other capabilities include deep-path traversals, parallel query execution, custom query modules in Python, Rust and C/C++, native Parquet and JSONL loading from local disk, S3 or HTTP, streaming ingestion from Kafka, Pulsar and RedPanda, an MCP server and integrations with LangChain, LlamaIndex and other agent frameworks, plus real-time schema introspection for Text2Cypher. Enterprise features cover high availability, multi-tenancy, label- and node-level role-based access control, SSO and 30+ granular permissions, encryption in transit, monitoring and backup/restore. Tooling includes Memgraph Lab for visual exploration and ad hoc queries, mgconsole, browser-based Playground sandboxes, Helm charts for Kubernetes and a managed Memgraph Cloud service on AWS.

Founding story

Memgraph was founded in 2016 by Dominik Tomicevic and Marko Budiselic. After consulting for corporate clients, the founders concluded that existing graph database systems were not engineered for performance, and spent roughly five years building an in-memory graph database from the ground up before releasing it broadly to developers.

Business model

Open-core software. A fully functional Community Edition is distributed free under the Business Source License, while an Enterprise Edition adds security, availability and operational features such as role-based access control, auditing, encryption, multi-tenancy, high availability and support SLAs. Memgraph Cloud provides managed Enterprise instances on AWS.

Revenue comes from commercial Enterprise Edition licensing (with support SLAs) and the managed Memgraph Cloud service, alongside a free Community Edition.

Traction

Reported production deployments include NASA (people knowledge graph for talent management), Cedars-Sinai (knowledge-driven machine learning in Alzheimer's research) and Capitec Bank, which scores more than 3.5 million clients daily with an end-to-end pipeline running in about two hours. The GitHub repository has roughly 4.6k stars, 284 forks and over 5,357 commits. The engine is documented to handle graph sizes of 100 GB to 4 TB and 1,000+ transactions per second for reads and writes.

Latest developments

Recent open-source releases include v3.13.0 (2026-09-09), which introduced a range of breaking changes across Cypher semantics, APOC compatibility, coordinator privileges and metrics defaults, and v3.13.1 (2026-09-14), which added experimental Enterprise-only FIPS 140-3 support via a dedicated Docker image and a --fips-mode flag, plus fixes to SHOW METRICS INFO. Product messaging now centres on serving as a graph engine for AI context, including GraphRAG, AI memory and agentic AI.