1. Free to run.
2. One line to install.

Free Edition is free to evaluate on your own hardware with 2 CPUs, native Cypher traversal, and exact vector search. Full source code is open on GitHub.

# Linux & macOS Quick Install curl -fsSL https://raw.githubusercontent.com/conwayek/GraphDB/main/scripts/install.sh | bash
# Windows PowerShell Quick Install irm https://raw.githubusercontent.com/conwayek/GraphDB/main/scripts/install.ps1 | iex

Supported platforms and release packages.

Official release binaries and packaging for all primary operating systems. Maintained and version-tracked via continuous integration.

Active Release Channel
LightningDB v1.2.0 Latest Stable
Access Paid Pro/Pro+ Binaries GitHub Releases

Ubuntu Linux

Free / Open

Ubuntu 20.04, 22.04, 24.04 LTS (x86_64 & aarch64)

Native Debian packaging with systemd service configuration. Pre-compiled with AVX2 & NEON.

Debian Linux

Free / Open

Debian 11 & 12 Bullseye / Bookworm (x86_64)

Standard tarball distribution or Debian package with zero runtime dependencies beyond GLIBC 2.31+.

macOS (Apple Silicon)

Free / Open

M1, M2, M3, M4 series (ARM64 macOS 12+)

Apple Disk Image (.dmg) and standalone command line binary with NEON SIMD hardware acceleration.

macOS (Intel x86_64)

Free / Open

Intel Core i7, i9, Xeon Mac hardware (macOS 11+)

Compiled with AVX2 SIMD vector extensions for accelerated cosine similarity and graph search.

Microsoft Windows

Free / Open

Windows 10, 11, Server 2022 (64-bit)

Windows Installer MSI package or portable ZIP with PowerShell automation and console client.

Pro & Pro+ (GPU Edition)

License Required

Linux x86_64 (CUDA 12.x / Tensor Core support)

Unlocks 16–32 cores, CUDA GPU graph traversal, entropy dynamic worker, and online single-IP lock.

Quickstart: serve your first GraphRAG query.

Launch the database with our built-in benchmark dataset and execute an end-to-end retrieval call in under 60 seconds.

STEP 01 Launch the Server

Start LightningDB with a 28,000-node academic co-authorship corpus pre-loaded into memory:

cd ~/graphdb
./build/graphdb serve --people 20000 --papers 8000 --port 7475

STEP 02 Call the GraphRAG Endpoint

Send a query to retrieve the 10 nearest papers and traverse 3 hops of related co-authors and citations:

curl -X POST http://localhost:7475/v1/rag \
  -H "Content-Type: application/json" \
  -d '{"query": "distributed consensus and vector search", "k": 10, "hops": 3}'

STEP 03 Inspect the Latency

The response includes e2e_ns timing metadata directly from the C++ engine (typically < 1 ms for this corpus):

{
  "hits": [...10 nearest documents...],
  "subgraph_nodes": 4200,
  "e2e_ns": 860000
}

Commercial Licensing.

Paid editions ship as an optimized binary and a cryptographic license bound to one machine. You send the hardware fingerprint. Pro adds CUDA GPU acceleration, entropy compaction, and 16 cores; Pro+ is 32 cores; Enterprise is unlimited and multi-server.

Request Commercial License Review Editions
  1. Generate a fingerprint

    Run the fingerprint utility on your target production host or container:

    graphdb license fingerprint
  2. Send it with your edition choice

    Provide your hardware key and desired core tier to our engineering sales team.

  3. Start with the license file

    Place the returned graphdb.license file in your database root directory and start the service.

Run the automated community proof.

Verify our benchmark claims against Neo4j Community on your own hardware using our reproducible Docker harness.

Step-by-Step Proof Guide Explore GitHub 
# Run proof community script
bash scripts/proof_community.sh