Knowledge Graph
The knowledge graph is the visual view of every page and link in your vault. Pages are nodes. Wikilinks are edges. Pages with the most connections are highlighted as hubs.
Where to find it
- macOS app — open it from the command palette (
Cmd+K→ Knowledge Graph) or from the View menu. - Web app — the underlying graph data is available via the API (
GET /vault/graph) and to AI clients via themindwiki_graphMCP tool. A canvas visualization lives in the macOS app today.
What the edges represent
The graph uses several edge types, weighted by signal strength:
| Edge type | Weight | When it fires |
|---|---|---|
| Wikilink | 5 | Page A contains [[Page B]] |
| Tag co-occurrence | 3 | Two pages share frontmatter tags |
| Title similarity | 3 | Page titles overlap meaningfully |
| Same folder | 2 | Pages live in the same folder |
| Same area | 1 | Pages share an area frontmatter value |
Higher weights pull connected pages closer in the force-directed layout. The result is a map where intentional links (wikilinks, shared tags) shape the structure more than incidental adjacency.
Hubs
A hub is a page with many inbound links — the kind of page that other pages reference often. Hubs are visually emphasized.
By default, MindWiki prefers pages whose titles or paths contain readme, index, overview, or schema when picking hub pages, but any page can become a hub if other pages link to it heavily.
Hubs are useful navigation: you can spot the high-leverage pages in your vault at a glance.
Reading the graph
A few patterns worth noticing when you look at your graph:
- Bridges. Edges that connect otherwise-distant clusters often represent the most original thinking — pages that link a research finding to a project, or a personal reflection to a framework.
- Orphans. Nodes with no connections at all are pages that haven't been linked to or from anything else. They might be valuable but invisible. The Vault Health view also surfaces orphans explicitly. See Vault Health & Permissions.
- Dense clusters. Tight groups of nodes are usually a single project, research area, or theme. If a cluster sprawls past your peripheral vision, you might be ready to split it into folders or hub pages.
Interacting with the graph (macOS)
- Click a node to open that page.
- Drag a node to reposition it. The layout settles back over time.
- Scroll wheel / pinch to zoom in and out.
- Drag empty space to pan.
The graph is cached so it renders instantly on subsequent opens. Edits to your vault refresh the cached graph during the next sync cycle.
Programmatic access
Any AI client connected through MCP can call the mindwiki_graph tool to retrieve the full graph as {nodes, edges}. Useful prompts:
- "Get my MindWiki graph and tell me which pages are hubs."
- "What clusters of related pages exist in my vault?"
- "Find pages that link to my onboarding project."
REST API access is also available with GET /vault/graph using a key with read scope. See API Access.
Where to go next
- Wikilinks & Backlinks — the connections that shape the graph
- Vault Health & Permissions — orphans and broken links
- Search & Views — other ways to browse your vault