GraphForge vs. Memgraph
What matters more: low-latency live graph processing, or analysis over alternative graph states?
Memgraph is an in-memory graph database server built for streaming and low-latency queries, with BFS and weighted shortest paths built into its Cypher dialect and the MAGE library of C++, Python, and Rust query modules. GraphForge is an embedded engine for analysis: openCypher plus 94 native algorithms in your process, persisted to Parquet, with checkpoints and recorded hypotheses. One is built to keep up with a changing graph; the other is built to reason carefully about a graph and keep a record of that reasoning.
| GraphForge | Memgraph | |
|---|---|---|
| Model | Embedded property graph | In-memory property graph |
| Version | 0.5.2 stable · 0.6.0-rc.52 | 3.13.2 (2026-10-09)17 |
| License | Apache-2.0 (since 0.5.0) | Business Source License 1.1 (core) / Memgraph Enterprise License (Enterprise features) |
| Status | Actively maintained2 | Actively maintained17 |
On this page
Should I choose Memgraph or GraphForge?
Memgraph for low-latency queries over a live, streaming graph; GraphForge for durable, in-process analysis.
Choose GraphForge when
- The analysis should run inside a notebook or application with no service to operate.
- The graph should persist compactly to disk as Parquet rather than live in RAM.
- You need to compare alternative explanations and keep checkpoints and provenance for each.
Choose Memgraph when
- Data arrives continuously from Kafka, Redpanda, or Pulsar and queries must reflect it immediately.
- Many clients need concurrent, low-latency access to one in-memory graph over Bolt.
- You want GPU-accelerated algorithms through MAGE's cuGraph integration, or dynamic algorithms with an Enterprise licence.
What can each one do?
20 of 26 capabilities differ. For 11, Memgraph's docs did not confirm support, so they are marked Not established rather than No.
Each capability has a status and a source. “Not established” means we could not confirm the feature from Memgraph's primary documentation. It does not mean Memgraph lacks it.
| Capability | GraphForge | Memgraph |
|---|---|---|
| Storage | ||
| Property graph model | Shipped Labeled property graph with typed properties. 1 | Yes Labelled property graph queried with Cypher over Bolt. 24 |
| Document model | Not offered No document-store model; properties accept nested, heterogeneous values. 1 | Not established Maps/lists are property values; no separate document model documented. |
| Relational tables | Not offered Not a relational database; results arrive as Arrow tables. 1 | Not established No relational model; RDBMS sources reached via migration modules. |
| Key-value access | Not offered No key-value API; data is a property graph. 1 | Not established No key/value API documented. |
| Columnar storage | Shipped Durable projects store graph data as Arrow and Parquet. 4 | Not established No columnar storage documented. |
| Vector search | Shipped Exact cosine retrieval over embeddings you supply; no ANN index. 5 | Yes USearch-backed vector indexes on nodes and edges; default metric l2sq, cosine available. 23 |
| Full-text search | Shipped Tantivy-backed BM25; hybrid text and vector retrieval through find(). 5 | Yes Tantivy-based text search; no longer experimental since 3.6. 20 |
| Geospatial | Not offered No spatial types or indexes are documented. | Yes Point index accelerates point.distance() and point.withinbbox() filters. 27 |
| Execution | ||
| Transactional workloads (OLTP) | Partial Local analytical scope; durable atomic writes with queued and optimistic writer modes. 6 | Yes Default IN_MEMORY_TRANSACTIONAL mode with WAL and snapshots. 19 |
| Analytical workloads (OLAP) | Shipped Analytical verbs and Cypher over a local columnar project. 7 | Yes IN_MEMORY_ANALYTICAL mode speeds import/analytics but drops ACID except manual snapshots. 19 |
| Graph traversal | Shipped Cypher MATCH patterns plus BFS, DFS, Dijkstra, A*, and reachability. 8 | Yes Built-in DFS, BFS, weighted shortest path, all shortest paths, K shortest paths. 21 |
| Variable-length paths | Shipped Bounded and unbounded patterns and shortestPath(). 8 | Yes Variable-length expansions ([*1..4]); quantified path patterns like --{1,4} unsupported. 24 |
| Graph algorithms | Shipped 94 native algorithms behind analysis verbs, not Cypher CALL procedures. 37 | Extension MAGE query-module library, shipped in memgraph-mage images/packages. 22 |
| Incremental computation | Not offered Algorithms recompute on request; no incremental mode is documented. 7 | Paid edition Dynamic (online) PageRank, betweenness, Katz, community detection require Enterprise since 3.0. 22 |
| Deployment | ||
| Embedded / in-process | Shipped In-process library; no server, daemon, or connection string. 9 | Not established No in-process embedding mode documented. |
| Server mode | Planned Optional remote peer extension planned after 0.6.0. 11 | Yes Standalone server speaking the Bolt protocol. 24 |
| WebAssembly | Not planned A browser WebAssembly engine was declined in ADR 0027. 12 | Not established No WebAssembly build documented. |
| Distributed cluster | Not planned Distributed clusters and HA replication are out of scope. 9 | Partial MAIN/REPLICA replication and Enterprise HA; replicas hold full copies, no sharding documented. 25 |
| Managed cloud | Not planned No hosted compute. The graphforge.sh Hub distributes repositories; execution stays local. 13 | Yes Memgraph Cloud on AWS, 6 regions, Enterprise instances up to 32 GB RAM. 26 |
| Investigation | ||
| Branching | Experimental Research versions, branches, slices, proposals, and forks ship in 0.6.0 release candidates. 14 | Not established No graph branching documented. |
| Snapshots / time travel | Shipped Named checkpoints you can diff and revert, plus transaction-time snapshots. 3 | Not established No point-in-time queries documented. |
| Immutable history | Shipped Knowledge ledger records immutable assertions and evidence lineage. 16 | Not established No append-only history documented. |
| Provenance | Shipped Provenance and knowledge ledger with evidence lineage. 16 | Not established No engine-level provenance documented. |
| Competing hypotheses | Shipped Immutable hypothesis groups with membership and selection events. 3 | Not established |
| Operations | ||
| ACID transactions | Partial Acknowledged-durable atomic publication; docs do not claim general ACID or serializable isolation. 6 | Yes ACID in transactional modes; IN_MEMORY_ANALYTICAL mode gives no ACID guarantees. 19 |
| Replication | Not planned HA replication is out of scope. 9 | Yes Community replication with manual failover; automatic failover via Raft coordinators is Enterprise. 25 |
Showing 26 of 26 capabilities.
How is each one built?
Memgraph: in-memory property graph. GraphForge: an embedded property graph library in Rust, persisted as Parquet.
GraphForge
GraphForge is a Rust library that runs inside your process: no daemon, server, or connection string. Projects persist to a local directory as Arrow and Parquet data plus JSON metadata, and every query and analysis verb returns an Apache Arrow table. Algorithms run in the same native engine as Cypher and are called as API verbs (rank, cluster, paths, analyze, similar), not as Cypher procedures.
Memgraph
Memgraph is a C++ server that keeps the graph in memory. It offers transactional and analytical in-memory modes plus an experimental on-disk mode; durability comes from write-ahead logs and snapshots. Traversal algorithms are part of the query engine, and other algorithms load as MAGE query modules.
Memgraph trades disk footprint for latency: the working graph sits in RAM on a server. GraphForge keeps the project on local disk in Arrow and Parquet and computes inside the calling process, so the cost model is a library call rather than a network round-trip to a resident service.
What the documentation says about Memgraph
- Memgraph offers three storage modes: IN_MEMORY_TRANSACTIONAL (default), IN_MEMORY_ANALYTICAL and ON_DISK_TRANSACTIONAL.19
- In the default transactional mode, durability comes from write-ahead logs and snapshots.19
- DFS, BFS, weighted shortest path, all shortest paths and K shortest paths are built into Memgraph without extra libraries, using syntax such as [*BFS] and [*KSHORTEST | 3].21
- Vector values are stored only in the USearch index backend while the property store keeps a reference, avoiding duplication.23
- Text search is built on the Tantivy engine and has not been experimental since Memgraph 3.6.20
- Memgraph can ingest from Kafka, Redpanda and Pulsar streams using CREATE ... STREAM queries and transformation modules.28
- Since Memgraph 3.0, dynamic (online) graph algorithms require a Memgraph Enterprise licence.22
- MAGE includes GPU-accelerated NVIDIA cuGraph integration plus NetworkX and igraph wrappers.22
- Memgraph Cloud is a managed AWS service offering Enterprise instances with up to 32 GB RAM and 8 CPU cores.26
Which one is faster?
Not measured yet. No head-to-head run has been published, so this page shows no performance chart.
Charts appear only after a reproducible run exists: pinned versions, disclosed hardware, generated data, raw results, and scripts.
No comparative results yet
The suites below apply to GraphForge and Memgraph. Each will be reported separately per benchmark class, never merged into one leaderboard.
Ingestion and writes
| Suite | Workloads | Metrics | Status |
|---|---|---|---|
| INGEST | CSV, JSONL, Parquet, graph edge lists | Records/s, MB/s, total elapsed | Not run |
| MUTATE | Insert, update, delete, upsert | Ops/s, p50/p95/p99 latency | Not run |
Query, analytics, and system
| Suite | Workloads | Metrics | Status |
|---|---|---|---|
| LOOKUP | ID lookup, indexed property filter | p50/p95/p99 latency | Not run |
| TRAVERSE | BFS, k-hop, variable-length path | Latency, visited edges/s | Not run |
| ALGO | PageRank, WCC, SCC, Louvain, SSSP | Time, memory, correctness | Not run |
| EMBED | Startup, load, reopen, binary size | ms, MiB, disk footprint | Not run |
| DURABLE | Restart, recovery, persistence | Recovery time, data integrity | Not run |
| CONCUR | Mixed readers and writers | Throughput, tail latency | Not run |
Which graph algorithms ship?
GraphForge: 94 native algorithms. Memgraph: 25 or more named in its docs, delivered as extension.
Algorithms are listed by name, as documented. Counts of extension modules or third-party libraries are not presented as native coverage.
GraphForge
94 Rust algorithms run in the engine behind five analysis verbs (rank, cluster, paths, analyze, similar) and return Arrow tables.37
A selection of the catalog:
- PageRank
- ArticleRank
- Betweenness (Brandes)
- Closeness
- Harmonic closeness
- Eigenvector
- HITS
- CELF influence maximization
- Triangle count
- K-core decomposition
- Louvain
- Leiden
- Label propagation
- Infomap
- Walktrap
- Girvan-Newman
- Weakly connected components
- Strongly connected components
- Dijkstra
- A*
- Bellman-Ford
- Yen's k-shortest paths
- Max flow / min cut
- Steiner tree
- Minimum spanning tree
- Topological sort
- Node2Vec
- GraphSAGE
- Fast random projection
- Node similarity (Jaccard)
Memgraph
Traversal algorithms are built into Cypher syntax; everything else comes from MAGE, a library of C++/Python/Rust query modules invoked with CALL.2122
Delivered as: extension. A selection of the catalog:
- Depth-first search
- Breadth-first search
- Weighted shortest path
- All shortest paths
- K shortest paths
- PageRank
- Betweenness centrality
- Degree centrality
- Katz centrality
- Louvain community detection
- Leiden community detection
- Weakly connected components
- Biconnected components
- Bridges
- Cycle detection
- Bipartite matching
- Graph coloring
- Max flow
- Node similarity
- k-means clustering
- Traveling salesman (tsp)
- Vehicle routing (vrp)
- node2vec
- GNN link prediction
- Temporal graph networks
Where can each one run?
GraphForge runs in-process only. Memgraph: embedded not established, server yes, cluster partial.
| Mode | GraphForge | Memgraph |
|---|---|---|
| Embedded / in-process | Shipped In-process library; no server, daemon, or connection string. 9 | Not established No in-process embedding mode documented. |
| Server mode | Planned Optional remote peer extension planned after 0.6.0. 11 | Yes Standalone server speaking the Bolt protocol. 24 |
| WebAssembly | Not planned A browser WebAssembly engine was declined in ADR 0027. 12 | Not established No WebAssembly build documented. |
| Distributed cluster | Not planned Distributed clusters and HA replication are out of scope. 9 | Partial MAIN/REPLICA replication and Enterprise HA; replicas hold full copies, no sharding documented. 25 |
| Managed cloud | Not planned No hosted compute. The graphforge.sh Hub distributes repositories; execution stays local. 13 | Yes Memgraph Cloud on AWS, 6 regions, Enterprise instances up to 32 GB RAM. 26 |
| Replication | Not planned HA replication is out of scope. 9 | Yes Community replication with manual failover; automatic failover via Raft coordinators is Enterprise. 25 |
What will my team write?
GraphForge: openCypher from Python, Node.js, Rust. Memgraph: Cypher (openCypher with Memgraph extensions).
Memgraph speaks Bolt and Cypher with its own extensions, such as [*BFS] and [*KSHORTEST], and extends through query modules written in C++, Python, or Rust. GraphForge uses standard openCypher, with algorithms called as Python, Node.js, or Rust verbs that return Arrow tables.
- Query languages
- GraphForge openCypher
- Memgraph Cypher (openCypher with Memgraph extensions)
- SDKs and bindings
- GraphForge Python, Node.js, Rust
- Memgraph Python, Rust, C#, Go, Java, JavaScript, Node.js, PHP
- Written in
- GraphForge Rust
- Memgraph C++
- Surfaces
- GraphForge Python, Node.js, Rust crates, CLI, VS Code extension, Notebooks, Agent skills
What are the trade-offs?
Memgraph keeps the graph in RAM and gates HA behind Enterprise. GraphForge has no streaming ingest or server mode.
A fair assessment cuts both ways.
GraphForge
- No streaming ingestion connectors; data arrives through Arrow, dataframes, or Cypher scripts.
- No server mode for concurrent remote clients today; a remote peer extension is planned after 0.6.0.
- No GPU-accelerated algorithms.
Memgraph
- The on-disk transactional storage mode is still experimental, and Memgraph recommends in-memory modes whenever possible.19
- IN_MEMORY_ANALYTICAL mode writes no WAL or periodic snapshots and gives no ACID guarantees besides manually created snapshots.19
- Automatic failover (high availability with Raft coordinators) is an Enterprise feature; Community replication requires manual failover.25
- Memgraph does not support Neo4j's NOT label expression or quantified path patterns such as --{1,4}.24
- Memgraph 3.13.0 made the cross_database module, including deprecated migrate.* aliases, Enterprise-only.18
How would I migrate?
Load Memgraph data through Arrow, pandas, or Polars. Standard openCypher carries over; procedure calls become verb calls.
Into GraphForge
- GraphForge 0.5.x loads data through Arrow, pandas, and Polars bulk calls or Cypher scripts; it has no native CSV loader.15
- Algorithms are API verbs (rank, cluster, paths, analyze, similar), not Cypher CALL procedures, so procedure calls become verb calls.7
- GraphForge passes the full openCypher 2024.3 TCK, so standard openCypher queries are the portable part of a migration.10
Moving from Memgraph
- Memgraph speaks Bolt and Cypher but documents differences from Neo4j, e.g. [*BFS] instead of shortestPath() and *KSHORTEST instead of SHORTEST k.24
- From 3.13.0, the cross_database module (and its deprecated migrate.* aliases) used for direct streaming migration requires an Enterprise licence.18
- Release 3.13.0 includes breaking Cypher semantic changes (character-based size(), octal literals, CALL {} scoping), so upgrades need query review.18
What else do people ask about Memgraph?
Is GraphForge faster than Memgraph? No head-to-head benchmark yet; GraphForge has no streaming connectors, but ships many MAGE-equivalent algorithms.
Is GraphForge faster than Memgraph?
We have not published a head-to-head benchmark. When we do, it will separate equivalent workloads, native optimized workloads, and end-to-end workloads, and include Memgraph's in-memory loading time.
Does GraphForge support streaming updates like Memgraph?
No. GraphForge has no Kafka or Pulsar connectors. Writes are durable atomic publications, and the engine is designed for analysis rather than continuous event ingestion.
Can I use Memgraph's MAGE algorithms in GraphForge?
Not as modules. Many of the same algorithms ship natively in GraphForge, including PageRank, Louvain, Leiden, betweenness, and weakly connected components, and are called as API verbs.
How was this researched?
From 28 primary sources, accessed 2026-10-19, each cited where it is used.
Capabilities were checked against primary documentation, repositories, and release notes, accessed 2026-10-19. Each cell records two things: the capability’s state and the evidence behind it. Every GraphForge cell carries a release status. Shipped means stable 0.5.2; experimental means 0.6.0 release candidates only. GraphForge is pre-1.0 and makes no backward-compatibility guarantee yet.
- Yes documented natively
- Partial limited or caveated
- Extension official extension or module
- Paid edition paid or enterprise edition only
- Removed removed, with the version cited
- Not established not confirmed by our research; not a “no”
- Shipped in stable 0.5.2
- Partial shipped with documented limits
- Experimental in 0.6.0 release candidates only
- Planned on the roadmap
- Not offered outside the product's model
- Not planned explicitly out of scope
Benchmark classes
- Equivalent workloads. The same logical operation with the same correctness requirement on every engine.
- Native optimized workloads. Each engine's best supported implementation, including projections, indexes, extensions, and native algorithms.
- End-to-end application workloads. Time from raw data to a usable result: import, indexing, projection, computation, and output retrieval.
Sources · accessed 2026-10-19
- GraphForge repository README — GitHub
- GraphForge releases — GitHub
- GraphForge 0.5.0 changelog — GitHub
- Storage architecture — GraphForge Book
- Analytics integration: search and retrieval — GraphForge Guide
- ADR 0018: acknowledged durability and isolation — GitHub
- Algorithm verbs — GraphForge Book
- Cypher guide — GraphForge Guide
- openCypher compatibility — GraphForge Reference
- openCypher TCK compliance — GraphForge Reference
- Remote authority peer extension (#215) — GitHub
- ADR 0027: native runtime boundary — GitHub
- Release scope and roadmap — GraphForge Docs
- Keep research history — GraphForge Guide
- Graph construction — GraphForge Guide
- Knowledge public API v1 — GraphForge Book
- Release v3.13.2 — memgraph/memgraph GitHub
- Release v3.13.0 — memgraph/memgraph GitHub
- Storage memory usage (storage modes) — Memgraph Docs
- Text search — Memgraph Docs
- Deep path traversal algorithms — Memgraph Docs
- Available advanced algorithms (MAGE) — Memgraph Docs
- Vector search — Memgraph Docs
- Differences in Cypher implementations — Memgraph Docs
- Replication — Memgraph Docs
- Memgraph Cloud — Memgraph Docs
- Indexes — Memgraph Docs
- Data streams — Memgraph Docs
Try GraphForge on your own graph.
Install the package, open a project, and run the same notebook we measure with.