obsidian-visualize — Claude Code command: vault graph as Canvas
The obsidian-visualize command maps your Obsidian vault visually: shows how notes connect, identifies hubs, and highlights isolated entries. Use it to see your second brain's shape at a glance.
MIT ★ 5,412 updated 2026-10-24
/plugin marketplace add eugeniughelbur/obsidian-second-brain
/plugin install obsidian-second-brain@obsidian-second-brainWhat obsidian-visualize does
The command scans your vault and exports a visual graph in Obsidian Canvas format. Instead of loading every note into context, it uses a scanner — this preserves token budget on large vaults. The graph displays nodes (notes) and edges (links between them), colored by type: entities in blue, projects in green, concepts in purple, daily notes in gray, sources in orange.
You can visualize the entire vault or focus on a specific note and its two-hop neighbourhood. Output includes a text report: top hubs by centrality (with link counts), bridges connecting clusters, orphan notes (no connections), and semantic link types (if defined via relations: in frontmatter). The graph saves as atlas.canvas in your vault root and opens directly in Obsidian.
Requires the obsidian-second-brain skill installed in your vault. All notes created or updated by this command must follow rules from references/ai-first-rules.md.
What it helps with
- visualize vault structure without manual diagramming
- find the most-linked notes (hubs through which knowledge flows)
- spot isolated entries that have no connections
- surface causal chains through typed relationships
- identify bridges that connect knowledge clusters
Example prompts for the agent
- Visualize my entire vault — show me the full knowledge graph
- Map the "Tide Gateway" project with all connected notes and ideas
- Build a Canvas showing only entities and their relationships
How to install obsidian-visualize
1Plugin marketplace
/plugin marketplace add eugeniughelbur/obsidian-second-brain/plugin install obsidian-second-brain@obsidian-second-brain2Manually: put the files in place
.claude/commands/obsidian-visualize.mdCall it with /obsidian-visualize. For all projects use ~/.claude/commands/
Manually: put the files in place
~/.codex/prompts/obsidian-visualize.mdCall it with /prompts:obsidian-visualize. Codex marks this format deprecated in favour of skills
Manually: put the files in place
.github/prompts/obsidian-visualize.prompt.mdManually: put the files in place
.gemini/commands/obsidian-visualize.tomlGemini stores commands as TOML: the prompt text goes into the prompt field
Questions about obsidian-visualize
How do I install obsidian-visualize in Claude Code?
Run these in order: /plugin marketplace add eugeniughelbur/obsidian-second-brain, then /plugin install obsidian-second-brain@obsidian-second-brain, then add .claude/commands/obsidian-visualize.md to your project.
What does obsidian-visualize do?
It scans your Obsidian vault, builds a graph of connections between notes, and exports it as an interactive Canvas. You see the vault's structure: where hubs sit, which notes are isolated, how clusters relate to each other.
Can I visualize just part of my vault?
Yes — pass the exact note title or alias as an argument, and the graph includes that note plus its two-hop neighbourhood. Without an argument, it visualizes the entire vault.
Source text
obsidian-visualize.md · MIT
---
description: Generate a visual canvas map of your vault - see the shape of your second brain and how knowledge connects
category: meta
triggers_en: ["visualize vault", "vault map", "canvas of vault", "show me the vault shape"]
triggers_de: ["Vault visualisieren", "Vault-Karte", "Vault als Canvas", "zeige mir die Struktur meines Vaults"]
triggers_es: ["visualiza el vault", "mapa del vault", "canvas del vault", "muéstrame la forma de mi vault"]
triggers_pt: ["visualize o vault", "mapa do vault", "canvas do vault", "mostre a forma do vault"]
triggers_zh: ["把知识库可视化", "生成知识图谱画布", "画出我的笔记关系", "看看知识库的整体结构"]
---
Use the obsidian-second-brain skill. Execute `/obsidian-visualize $ARGUMENTS`:
The optional argument is a scope: an EXACT note title or alias (a project, an entity), or "full" for the entire vault. Default: full vault. A free-form topic phrase will not match - resolve it to a real note title first (e.g. via /obsidian-find).
1. Read `_CLAUDE.md` first if it exists in the vault root
2. Build the graph deterministically with the scanner instead of reading every note into context (a full-vault read is O(read-everything) and burns the budget). Run from the skill root (its absolute path was given at session start as **Skill root**; substitute it for `SKILL_ROOT`):
```bash
uv run --directory "SKILL_ROOT" scripts/link_graph.py --path "<vault>" [--scope "<topic/project/entity>"]
```
It returns JSON with `nodes` (path, title, `type`, folder, in/out/`degree`), `edges` (resolved `[[wikilink]]` pairs, each `{from, to, type: "link"}`), `typed_edges` (the semantic overlay from `relations:` frontmatter - `{from, to, type}` where type is `supersedes`, `depends_on`, `caused`, etc. per `references/ai-first-rules.md` Rule 6), `typed_edge_problems` (unhonored typed edges), and `stats` (`node_count`, `edge_count`, `typed_edge_count`, `orphan_count`, `dangling_link_count`, `top_hubs`, `orphans`). Pass `--scope` with an exact note title/alias (the script keeps that note plus its 2-hop neighborhood); omit it for the full vault. If the result has `node_count: 0`, the scope did not resolve to a note - find the real title and rerun. Use this JSON as the graph - only open individual notes if you need a label the scan did not provide.
3. Generate a JSON Canvas file (`.canvas`) compatible with Obsidian's native canvas viewer:
Structure:
```json
{
"nodes": [
{"id": "1", "type": "file", "file": "wiki/entities/Ada Lovelace.md", "x": 0, "y": 0, "width": 250, "height": 60},
{"id": "2", "type": "file", "file": "wiki/projects/Tide Gateway.md", "x": 300, "y": 0, "width": 250, "height": 60}
],
"edges": [
{"id": "e1", "fromNode": "1", "toNode": "2"}
]
}
```
Layout rules:
- **Hub nodes** (most links) go in the center, larger
- **Cluster by type**: entities on the left, projects top-right, concepts bottom-right, daily notes bottom
- **Color by type**: entities = blue, projects = green, concepts = purple, daily = gray, sources = orange
- **Edge labels** = when two nodes connect through multiple links, record the count in the edge `label` (JSON Canvas edges have no thickness property). When a `typed_edges` entry connects the same pair, label the edge with the relation type instead (e.g. `supersedes`, `depends_on`) - a typed edge is more informative than a count
- **Orphan nodes** placed at the edges with a red border (easy to spot)
4. Save to vault root as `atlas.canvas` (or `atlas-{topic}.canvas` if scoped)
5. Also generate a text summary with centrality ranking (use the scanner's `stats` and per-node `degree`/`in`/`out` directly - do not recompute by hand):
- Total nodes and edges (`stats.node_count`, `stats.edge_count`), plus the `dangling_link_count` (wanted notes - links to unwritten notes).
- **Typed edges** - if `stats.typed_edge_count` > 0, summarize the semantic overlay: count by relation type (from `typed_edges`), name the longest `supersedes`/`depends_on`/`caused` chains (these are the reasoning paths flat links can't express), and flag any `typed_edge_problems` as a nudge to run `/obsidian-health` for the full lint. If it's 0, note that the vault has no typed edges yet and point to `references/ai-first-rules.md` Rule 6 as the way to add them.
- **Hub nodes (centrality)** - top 5 from `stats.top_hubs` (already ranked by degree), each with its link count and a one-line "everything flows through this because..." note. A hub qualifies if its degree is at least 3x the median, or it sits in the top 1% of the vault - whichever surfaces fewer.
- **Bridge nodes** - nodes that, if removed, would split a cluster. Rank by betweenness (approximate: count the shortest paths each node sits on between the top-10 hubs). These are the load-bearing connectors; surface the top 3 with the two clusters each one joins.
- **Orphan nodes** - no connections, listed by type. Flag any that are >30 days old (stale orphans are higher-priority cleanup targets than fresh ones).
- **Clusters** - groups of tightly connected notes, named by their hub. Note any cluster with <3 cross-cluster edges (those are silos).
- **Centrality skew** - if one node holds >25% of total edges, call it out as a single point of failure for navigation.
6. Append to the operation log: if `Logs/` exists write `**HH:MM** - visualize | Canvas generated - X nodes, Y edges, Z orphans` to `Logs/YYYY-MM-DD.md`; otherwise append `## [YYYY-MM-DD] visualize | Canvas generated - X nodes, Y edges, Z orphans` to `log.md`
The user can open the `.canvas` file in Obsidian to visually explore their vault's knowledge graph.
---
**AI-first rule:** Every note created or updated by this command MUST follow `references/ai-first-rules.md` - `## For future agent` preamble, rich frontmatter (`type`, `date`, `tags`, `ai-first: true`, plus type-specific fields), recency markers per external claim, mandatory `[[wikilinks]]` for every person/project/concept referenced, sources preserved verbatim with URLs inline, and confidence levels where applicable. If that path does not resolve from your working directory, search upward for it; if you still cannot read it, say so before writing rather than producing a note that silently skips the rule. The vault is for future agent retrieval - not human reading.
**Anti-fabrication:** Search exhaustively before claiming any note, person, or file is absent - false absence is the most common failure mode - and never invent facts, entities, or dates (mark unknowns as `TBD`). See the anti-fabrication and search-completeness hard rules in `references/ai-first-rules.md`.
© Eugeniu Ghelbur · MIT · eugeniughelbur/obsidian-second-brain