Truespar Traverse
A graph database built for speed. Traverse supports Cypher and the full ISO GQL standard, works with existing Neo4j drivers and tools, and runs as a server, embedded in your application, or entirely in a browser tab.
In-Memory Speed
The whole graph lives in memory. Across our 33-query benchmark, Traverse holds a 7.3x geometric mean speedup over Neo4j, and deep traversals run up to 376x faster.
Works With Your Neo4j Stack
Bolt v5 and v6, Cypher, the complete APOC core catalog and GDS-style algorithms are all built in. Existing drivers and most tools connect unchanged.
Runs Anywhere
A standalone server, an embedded library for .NET, Python, Java, Node.js and Go, or the full engine compiled to WebAssembly in a browser tab.
Use It Your Way
Server Mode
Run Traverse as a standalone service and connect over Bolt, HTTP or gRPC. Multiple databases per server, API key authentication, TLS. Studio ships inside the same binary.
Embedded Mode
Add Traverse as a library and run queries in-process, with no server to manage. Available for .NET, Python, Java, Node.js, Go and C.
In the Browser
The full engine runs in a browser tab via WebAssembly, with databases persisted locally. Try it at traverse.truespar.com, or embed it in your own app with the npm package.
Cypher, GQL and APOC
Cypher
Traverse implements the Cypher 25 surface: quantified path patterns, dynamic labels, CALL IN TRANSACTIONS, vector similarity inside MATCH, and more. If you know Cypher from Neo4j, you already know Traverse.
Full ISO GQL
100% conformance with the ISO/IEC 39075 GQL standard, the first international standard database language since SQL. Pick the dialect per query, in Studio or over any protocol.
APOC Built In
The complete APOC core catalog, 246 functions and 190 procedures, ships in the binary. Loaders, exports, triggers, background jobs. Nothing to install, nothing to allowlist.
Analytics and ML Built In
Graph Algorithms
36+ algorithms callable from Cypher: PageRank, Louvain, shortest paths, similarity, embeddings and more. In a single-thread comparison on the Pokec dataset, Traverse won 17 of 17 against Neo4j GDS.
ML Pipelines
Train node-classification models, including graph neural networks, entirely from Cypher. Training runs on GPU or CPU, models persist in the database, and Studio has a panel for the whole pipeline.
Vector & Full-Text Search
An HNSW vector index with quantization, BM25 full-text search, and text and point indexes. All of it persists in the database file and participates in crash recovery.
Built for Production
ACID Transactions
Every write is transactional. Readers never block writers, and writers never block readers.
Durability, Your Choice
Each database is one portable .tvdb file, compressed and checksummed. The default is snapshot persistence; turn on write-ahead logging and every committed write survives a hard crash.
Multi-Database
Host several databases on one server. Create, drop and switch between them from Studio or with Cypher commands.
See the numbers
We benchmark against Neo4j and Memgraph across 33 queries and publish the results: a 7.3x geometric mean speedup over Neo4j, up to 376x on deep traversals.