GraphForge vs. Neo4j
How do mature Cypher and Graph Data Science workflows compare with an embedded graph engine?
Neo4j is the reference server-based property graph database, with the largest Cypher ecosystem and a Graph Data Science library that projects data into memory before running algorithms. GraphForge runs openCypher and 94 native algorithms inside your own process, persists to Parquet, and returns Arrow tables. The choice is mostly about where the graph lives: in a shared database service, or beside the notebook, script, or application doing the analysis.
On this page
Should I choose Neo4j or GraphForge?
Neo4j for a shared, always-on graph service; GraphForge for analysis that runs inside your own process.
Choose GraphForge when
- The graph belongs to one analysis, notebook, or application and should not need a server.
- You want algorithm results as Arrow tables without a projection step or a write-back mode.
- Your investigation needs checkpoints, provenance, and recorded competing hypotheses next to the data.
Choose Neo4j when
- Many applications and users need one shared, always-on graph database with role-based access control.
- You depend on the Neo4j ecosystem, such as APOC, Bloom, drivers in five languages, or Aura managed hosting.
- You need clustering, online backup, or sharding, and can license Enterprise Edition or Infinigraph.
What can each one do?
20 of 26 capabilities differ. For 11, Neo4j'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 Neo4j's primary documentation. It does not mean Neo4j lacks it.
| Capability | GraphForge | Neo4j |
|---|---|---|
| Storage | ||
| Property graph model | Shipped Labeled property graph with typed properties. 1 | Yes Native graph storage and processing in all editions. 17 |
| Document model | Not offered No document-store model; properties accept nested, heterogeneous values. 1 | Not established No document model documented; maps/lists are property values. |
| Relational tables | Not offered Not a relational database; results arrive as Arrow tables. 1 | Not established No relational model documented. |
| 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; block store format is row/graph oriented. |
| Vector search | Shipped Exact cosine retrieval over embeddings you supply; no ANN index. 5 | Yes Lucene HNSW vector indexes, cosine/euclidean, 1-4096 dims, all editions; native VECTOR type needs block format (EE/Aura). 24 |
| Full-text search | Shipped Tantivy-backed BM25; hybrid text and vector retrieval through find(). 5 | Yes Full-text node and relationship indexes in all editions. 17 |
| Geospatial | Not offered No spatial types or indexes are documented. | Partial POINT type (WGS-84/Cartesian, 2D/3D) with point indexes; no polygons or lines documented. 26 |
| Execution | ||
| Transactional workloads (OLTP) | Partial Local analytical scope; durable atomic writes with queued and optimistic writer modes. 6 | Yes ACID transactional database with cost-based optimizer in all editions. 17 |
| Analytical workloads (OLAP) | Shipped Analytical verbs and Cypher over a local columnar project. 7 | Extension Graph analytics via GDS plugin; Parallel Cypher runtime is Enterprise-only. 1719 |
| Graph traversal | Shipped Cypher MATCH patterns plus BFS, DFS, Dijkstra, A*, and reachability. 8 | Yes Cypher pattern matching over native graph storage. 17 |
| Variable-length paths | Shipped Bounded and unbounded patterns and shortestPath(). 8 | Yes GQL-conformant quantified path patterns plus legacy variable-length relationships. 27 |
| Graph algorithms | Shipped 94 native algorithms behind analysis verbs, not Cypher CALL procedures. 37 | Extension Graph Data Science library plugin; Community GDS capped at 4 CPU cores. 19 |
| Incremental computation | Not offered Algorithms recompute on request; no incremental mode is documented. 7 | Not established GDS runs on in-memory projections; no incremental algorithm maintenance documented. |
| Deployment | ||
| Embedded / in-process | Shipped In-process library; no server, daemon, or connection string. 9 | Yes Embeddable in JVM applications via DatabaseManagementServiceBuilder (org.neo4j:neo4j). 22 |
| Server mode | Planned Optional remote peer extension planned after 0.6.0. 11 | Yes Server with Bolt protocol, Cypher Shell and Browser. 17 |
| 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 | Paid edition Clustering is Enterprise; sharded property databases require Infinigraph subscription. 1725 |
| Managed cloud | Not planned No hosted compute. The graphforge.sh Hub distributes repositories; execution stays local. 13 | Yes Neo4j Aura: AuraDB, AuraDS and Aura Graph Analytics. 23 |
| Investigation | ||
| Branching | Experimental Research versions, branches, slices, proposals, and forks ship in 0.6.0 release candidates. 14 | Not established No database branching documented. |
| Snapshots / time travel | Shipped Named checkpoints you can diff and revert, plus transaction-time snapshots. 3 | Not established No point-in-time query documented. |
| Immutable history | Shipped Knowledge ledger records immutable assertions and evidence lineage. 16 | Not established No append-only history; Change Data Capture (Enterprise) exposes change events. |
| 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-compliant transactions in all editions. 17 |
| Replication | Not planned HA replication is out of scope. 9 | Paid edition Enterprise clusters: primaries (majority writes) plus secondaries for read scaling. 25 |
Showing 26 of 26 capabilities.
How is each one built?
Neo4j: native 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.
Neo4j
Neo4j is a Java database server with a native property-graph store and Cypher runtimes; Enterprise adds the pipelined and parallel runtimes, clustering, and the block store format. Graph Data Science is a separate plugin: data is projected from the database into an in-memory graph catalog, algorithms run as Cypher procedures, and results stream back, mutate the projection, or are written to the database.
The main architectural difference is the projection boundary. In Neo4j, analytics run on an in-memory copy of the stored graph, managed through a catalog. In GraphForge, algorithms read the same local project that queries do, and every result is an Arrow table your code already holds.
What the documentation says about Neo4j
- Since 2025.01, Neo4j server uses calendar versioning in the format YYYY.MM.Patch, with some releases designated LTS.17
- Neo4j ships as GPLv3 Community Edition, Enterprise Edition, and Infinigraph, an Enterprise variant with automatic sharding that needs a separate subscription.17
- GDS must load data from the database into an in-memory graph catalog via projections before running algorithms.19
- GDS algorithms run in stream, stats, mutate (write back to the projection) or write (persist to the database) execution modes.20
- GDS Community Edition includes all algorithms but limits concurrency to 4 CPU cores and the model catalog to 3 models.19
- Neo4j can be embedded in a Java application by building a DatabaseManagementService from the org.neo4j:neo4j dependency.22
- Vector indexes are powered by Apache Lucene's HNSW implementation and are available in both Community and Enterprise editions.24
- A clustered database stays writable while a simple majority of its primary allocations is running, and clients are guaranteed to read their own writes.25
- Neo4j Aura offers AuraDB tiers Free, Professional, Business Critical and Virtual Dedicated Cloud, plus AuraDS and Aura Graph Analytics.23
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 Neo4j. 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 |
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 |
| JOIN | Multi-relation pattern matching | Latency, rows/s | Not run |
| ALGO | PageRank, WCC, SCC, Louvain, SSSP | Time, memory, correctness | Not run |
| HYBRID | Vector + graph + metadata filters | Recall@k, latency, throughput | Not run |
| EMBED | Startup, load, reopen, binary size | ms, MiB, disk footprint | Not run |
| SCALE | Increasing graph size and density | Throughput degradation, limits | Not run |
Which graph algorithms ship?
GraphForge: 94 native algorithms. Neo4j: 25 or more named in its docs, delivered as plugin.
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)
Neo4j
Analytics come from the Graph Data Science (GDS) library: project data into an in-memory graph catalog, then run algorithms as Cypher procedures in stream, stats, mutate or write mode.1921
Delivered as: plugin. A selection of the catalog:
- PageRank
- ArticleRank
- Betweenness Centrality
- Closeness Centrality
- Harmonic Centrality
- Eigenvector Centrality
- HITS
- CELF (influence maximization)
- Louvain
- Leiden
- Label Propagation
- Weakly Connected Components
- Strongly Connected Components
- Triangle Count
- Local Clustering Coefficient
- K-Core Decomposition
- HDBSCAN
- Node Similarity
- K-Nearest Neighbors
- Dijkstra Source-Target
- A*
- Yen's K Shortest Paths
- Bellman-Ford
- Max Flow
- Topological Sort
Where can each one run?
GraphForge runs in-process only. Neo4j: embedded yes, server yes, cluster paid edition.
| Mode | GraphForge | Neo4j |
|---|---|---|
| Embedded / in-process | Shipped In-process library; no server, daemon, or connection string. 9 | Yes Embeddable in JVM applications via DatabaseManagementServiceBuilder (org.neo4j:neo4j). 22 |
| Server mode | Planned Optional remote peer extension planned after 0.6.0. 11 | Yes Server with Bolt protocol, Cypher Shell and Browser. 17 |
| 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 | Paid edition Clustering is Enterprise; sharded property databases require Infinigraph subscription. 1725 |
| Managed cloud | Not planned No hosted compute. The graphforge.sh Hub distributes repositories; execution stays local. 13 | Yes Neo4j Aura: AuraDB, AuraDS and Aura Graph Analytics. 23 |
| Replication | Not planned HA replication is out of scope. 9 | Paid edition Enterprise clusters: primaries (majority writes) plus secondaries for read scaling. 25 |
What will my team write?
GraphForge: openCypher from Python, Node.js, Rust. Neo4j: Cypher (Cypher 25 / Cypher 5).
Both speak Cypher. GraphForge passes the full openCypher 2024.3 TCK, but it has no CALL procedures: algorithms and search are API verbs in Python, Node.js, or Rust rather than gds.* procedures. Neo4j queries that use APOC, GDS, or other procedures need to be rewritten as verb calls. Neo4j developers work through Bolt drivers against a server; GraphForge developers import a package.
- Query languages
- GraphForge openCypher
- Neo4j Cypher (Cypher 25 / Cypher 5)
- SDKs and bindings
- GraphForge Python, Node.js, Rust
- Neo4j .NET, Go, Java, JavaScript, Python
- Written in
- GraphForge Rust
- Neo4j Java
- Surfaces
- GraphForge Python, Node.js, Rust crates, CLI, VS Code extension, Notebooks, Agent skills
What are the trade-offs?
Neo4j gates clustering and RBAC behind Enterprise. GraphForge has no server, no CALL procedures, and is pre-1.0.
A fair assessment cuts both ways.
GraphForge
- No server, clustering, or role-based access control; GraphForge is a single-machine, in-process engine.
- No CALL procedures, so APOC and GDS procedure calls do not port directly.
- Vector search is exact rather than approximate, and you supply the embeddings.
- Pre-1.0, with no backward-compatibility guarantee yet.
Neo4j
- Community Edition is aimed at single-instance deployments; clustering, online backup, multiple databases and role-based access control are Enterprise-only.17
- The Pipelined and Parallel Cypher runtimes, block store format and Change Data Capture are not available in Community Edition.17
- Vector search is approximate, only one vector index can exist per label/type and property, and same-transaction changes are not visible to the index.24
- GDS algorithms run on a projected in-memory copy, so results reach the database only in write mode.20
- Spatial support covers POINT values only; points in different coordinate reference systems cannot be compared.26
How would I migrate?
Load Neo4j 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 Neo4j
- neo4j-admin database import full bulk-loads CSV or Parquet input into a new database offline.28
- Legacy variable-length relationship syntax still works but is not GQL-conformant; quantified path patterns are the portable form.27
- Embeddings stored as LIST<FLOAT> properties work in all editions, but native VECTOR properties need block format, which only Enterprise and Aura offer.24
What else do people ask about Neo4j?
Is GraphForge a Neo4j alternative? Yes for local analysis: plain Cypher carries over and GDS algorithms map to native verbs.
Is GraphForge a Neo4j alternative?
For local analysis, yes: it runs openCypher and graph algorithms in-process without a server. It is not a replacement for a shared Neo4j deployment that serves many applications, needs clustering, or relies on access control.
Can GraphForge run my Neo4j Cypher queries?
Standard openCypher usually carries over, since GraphForge passes the full openCypher 2024.3 TCK. Queries that call procedures (CALL apoc.*, CALL gds.*) do not, because GraphForge exposes algorithms and search as API verbs instead.
Does GraphForge have an equivalent of Neo4j Graph Data Science?
GraphForge ships 94 native algorithms, including PageRank, Louvain, Leiden, betweenness, and Node2Vec, behind rank, cluster, paths, analyze, and similar verbs. There is no projection catalog: algorithms run on the local project and return Arrow tables.
How do I move data from Neo4j to GraphForge?
Export nodes and relationships to CSV or Parquet, load them into pandas, Polars, or Arrow, and publish them into a GraphForge project with the bulk APIs. GraphForge has no native CSV loader, so the dataframe step is required.
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
- Introduction (editions and versioning) — Neo4j Operations Manual
- Tags (2026.09.0) — neo4j/neo4j GitHub
- Introduction — Neo4j Graph Data Science
- Running algorithms — Neo4j Graph Data Science
- GDS algorithm documentation pages — neo4j/graph-data-science GitHub
- Embedding Neo4j in your Java application — Neo4j Java Reference
- Neo4j Aura overview — Neo4j Aura Docs
- Vector indexes — Neo4j Cypher Manual
- Introduction: Neo4j clustering architecture — Neo4j Operations Manual
- Spatial values — Neo4j Cypher Manual
- Variable-length patterns — Neo4j Cypher Manual
- Full import — Neo4j Operations Manual
Try GraphForge on your own graph.
Install the package, open a project, and run the same notebook we measure with.