Open Source — MIT License

Visualize any knowledge graph.
In 3D.

Interactive 3D graph visualization with pathway-aware layouts, flow particles, molecule viewer, and analytics. Built with Three.js — zero dependencies.

Get the Plugin See it in action

> /plugin marketplace add cdeust/neural-graph-visualizer
> /plugin install neural-graph-visualizer

Requires Node.js 18+ and a WebGL-capable browser

Psoriasis immunopathology knowledge graph with cascade layout showing immune cells, cytokines, chemokines, and signaling pathways as glowing 3D nodes with flow particles — Cascade layout — Psoriasis immunopathology pathway with 475 synapses across 8 node types

Neural Graph molecule viewer showing Adalimumab (Humira) protein structure in ball-and-stick mode with 4466 atoms and 3 chains — Molecule viewer — Adalimumab (Humira) rendered from PDB with 4,466 atoms

5 Layout Strategies

12 MCP Tools

6 Node Types

0 Dependencies

Not just a graph. A reasoning tool.

Topology-driven layouts that reveal structure. Molecule viewers that show what the nodes actually are. Analytics that surface what matters.

◉

Pathway-Aware Layouts

Nodes positioned by topology, not physics. Five strategies — cascade, pipeline, radial, force, and auto — computed from directed edge types and hierarchical ranks.

⚖

Molecule Viewer

Nodes with PDB IDs get a 3D protein viewer (ribbon, ball-and-stick, surface). Nodes with SMILES get 2D/3D molecule rendering. Built for biotech and pharma.

✨

Flow Particles & Bloom

Synaptic pulses travel along edges. Cinematic bloom on all nodes. Holographic grid, 4,000 ambient dust particles. Visually rich without external assets.

☰

Analytics Dashboard

Node and edge counts, type distribution, stage filtering, top accessed files. Toggle with keyboard shortcut A. Filterable and searchable in real time.

⚙

MCP Integration

12 tools for Claude Code — open visualization, search nodes, import JSON/CSV, manage threads, add cross-references, query the brain index. Full programmatic access.

★

Research Templates

Drug discovery pipeline, blank starter. Copy a template, edit the JSON, launch. Industry presets for healthcare, pharma, food safety — or bring your own data.

Five layout strategies. Auto-detected.

The layout engine classifies your edges and picks the best strategy. Or you choose.

Strategy	Axis	Best For	Auto-Detected When
Cascade	Y (top → bottom)	Signaling pathways, dependency chains	>40% flow-type edges
Pipeline	X (left → right)	Sequential stages, workflows	Contains `proceeds_to` edges
Radial	Center → out	Target-centric views, interaction networks	Manual selection
Force	Physics-based	Generic graphs, exploration	Default fallback
Auto	Detected	Any — picks from edge distribution	Default when omitted

12 MCP tools. Full programmatic access.

Control the graph from Claude Code. Import data, search, add cross-references, manage threads — all via natural language.

open_visualization

get_graph

get_stats

import_json

import_csv

reindex_brain

get_brain_index

update_brain_entry

manage_thread

add_cross_reference

remove_cross_reference

> /neural-graph
# or ask Claude naturally:
> "Open the neural graph visualization"
> "Import my research data from data.json"

One JSON file. Three sections.

Config, nodes, edges. Define node types with colors and shapes, then describe your graph.

Config & Nodes

"_config": {
  "name": "My Graph",
  "nodeTypes": {
    "concept":  { "color": "#45aaf2", "shape": "sphere" },
    "entity":   { "color": "#26de81", "shape": "hex" }
  },
  "_layout": { "strategy": "auto" }
},
"nodes": [{
  "id": "node_1",
  "name": "My Node",
  "type": "concept",
  "pdbId": "5DK3"
}]

Edges & Edge Types

"edges": [{
  "source": "node_1",
  "target": "node_2",
  "weight": 0.8,
  "edgeType": "activation"
}]

// Flow edges (drive layout):
// activation, production,
// differentiation, proceeds_to,
// validates, causes

// Inhibition edges (flank targets):
// inhibition

// Proximity edges (no layout effect):
// binding, expression, synergy,
// relates_to, participates_in

Built for research. Works for anything.

Any directed graph with typed nodes and edges. Drop in your JSON and go.

⚙

AI Pipeline Runs

Visualize every stage, artifact, and agent decision from your autonomous pipeline as a live neural graph.

⚛

Drug Discovery

Map targets, compounds, pathways, and clinical endpoints. PDB protein viewer and SMILES rendering built in.

⚖

Food Safety Networks

Trace contamination pathways, regulatory dependencies, and supply chain relationships across facilities.

♦

Any Knowledge Graph

Codebase architecture, org charts, research citation networks, threat models. If it has nodes and edges, it renders.

Get started in 2 commands

Install as a Claude Code plugin or run standalone from CLI.

Claude Code Plugin

          # Install:

          > /plugin marketplace add cdeust/neural-graph-visualizer

          > /plugin install neural-graph-visualizer

          # Launch:

          > /neural-graph

Standalone CLI

          $ git clone https://github.com/cdeust/neural-graph-visualizer.git

          $ cd neural-graph-visualizer

          $ ./scripts/setup.sh

          # Launch with your data:

          $ node scripts/launch.js your-data.json

Free & Open Source

MIT licensed. Clone it, fork it, ship it. Zero dependencies, zero vendor lock-in.

Star on GitHub See the Pipeline

Visualize any knowledge graph.In 3D.