Memgraph

Memgraph is an in-memory graph database developed by Memgraph Ltd. (founded 2016, first 1.0 release in 2020). Written in C++, it stores a labeled property graph — nodes and relationships each carrying key/value properties — and is engineered for low-latency, high-throughput workloads such as fraud detection, network and supply-chain analysis, identity/access management, knowledge graphs, and, more recently, GraphRAG and agentic-AI "memory" backends. It keeps the working graph resident in RAM by default (with an optional on-disk storage mode for larger-than-memory datasets), provides ACID transactions with strong consistency, and offers Raft-based replication with automatic failover in its enterprise tier.

Its query surface is deliberately Neo4j-compatible: Memgraph speaks Cypher over the Bolt protocol, so most drivers, client libraries, and tooling built for Neo4j work with little or no change, and it is an active contributor to the openCypher standardization effort. This positioning makes it primarily a migration target and drop-in alternative rather than a new data model. The source is available under the Business Source License 1.1 (a source-available, time-delayed license), not a classic OSI open-source license, with a separate Memgraph Enterprise License for advanced features — a distinction worth noting for anyone assuming full open-source freedoms.

For the format-conversion and graph-tooling world, Memgraph's relevance is mostly at the ingestion and analytics boundaries. It imports tabular data via the Cypher LOAD CSV clause (and reads Parquet, ORC, and Arrow/Feather through PyArrow), and can consume live streams from Kafka, Pulsar, and Redpanda to update the graph in real time. The bundled MAGE (Memgraph Advanced Graph Extensions) library exposes 40+ algorithms — PageRank, community detection, node2vec embeddings, GNN-based link prediction — as callable Cypher procedures, with integrations to NetworkX, igraph, and cuGraph. Custom query modules can be written in Python, C/C++, or Rust, and the Python OGM/query-builder library GQLAlchemy eases application integration.

Memgraph Lab, its visual client, renders graphs (via the Orb library), runs ad-hoc Cypher, and supports natural-language querying. Honest limitations: RAM-bound sizing makes very large graphs costly, the ecosystem is smaller than Neo4j's, and it lacks native readers for interchange formats like GraphML or GML — data must arrive as CSV, a supported stream, or Cypher statements.