Graph
The graph is the relationship layer behind everything else in SubImage: the same normalized resources you browse in Inventory, connected by the edges Cartography discovers. The Graph explorer lets you work with those connections directly, tracing how resources relate rather than reading them as flat tables.
The Graph explorer
Open Graph to explore the relationship graph. This is a purpose-built visual explorer (currently in beta), not a generic database browser: it renders resources as an interactive node graph and lets you expand and trace connections.
You drive it with Cypher from the query bar, which has schema-aware autocomplete scoped to your enabled modules. Results come back as a graph when your query returns nodes and relationships, or as a table for scalar queries like RETURN count(*); a switch toggles between the two.
Once nodes are on screen you can:
- Click to select; double-click to pull in a node's neighbors; triple-click to grab the whole connected component.
- Right-click a node to expand specific relationship types, change the layout, or focus on it.
- Highlight the shortest path between two selected nodes.
The explorer runs queries through your authenticated SubImage session, so it needs no separate database credentials. It is the same view the Visualize and Explore buttons across the product open into.
You rarely need to write Cypher from scratch. Start from a resource in Inventory or a CVE, click Explore or Visualize, and expand outward from there. When you do need a query, the subimage-mcp:build-cypher-query skill will write and validate one for you.
Direct Neo4j access
The in-app explorer covers most needs. If you need to query the graph from the external Neo4j Browser or a script, see Neo4j API Keys for dedicated read-only database credentials.
Programmatic access
The SubImage MCP server does not require separate Neo4j credentials. Its subimageRunCypher tool runs read-only Cypher against the graph, and the schema tools (subimageListModuleSchemaNodes, subimageGetNodesSchema) help an assistant discover what to query. See Connect via MCP.