Skena — Visual Research Canvas for VS Code

Render JSON Canvas (.canvas) files as interactive node graphs right inside VS Code. Preview markdown notes, notebooks, code, and charts side-by-side. Navigate with vim keys. Works over Remote SSH. Obsidian-compatible.

Screenshots

A full research canvas — notebook charts, correlation matrices, backtest results, and code all on one spatial board.

Navigate your research by relationship, not by folder. Nodes auto-update when files change on disk.

Write notes alongside the artifacts they describe. Monaco editor with full vim bindings — i, o, yy, dd, p all work.

Pin notebook cell outputs as standalone nodes. Drag, zoom, connect — then save and push to git.

Why Skena

Quantitative research (or any deep technical work) generates a web of interconnected files — notebooks, results, code, configs, notes. Most tools force a choice:

Skena brings the spatial canvas experience into VS Code, so it works wherever VS Code works — including over Remote SSH on headless servers.

It reads and writes the standard JSON Canvas 1.0 format, so .canvas files are fully compatible with Obsidian.

Features

Rich inline previews

• Markdown (.md) — rendered with frontmatter header bar, status badges, scrollable content
• Jupyter Notebooks (.ipynb) — code cells, markdown cells, chart outputs, base64 images
• Python / YAML — syntax-highlighted code preview via Shiki
• Images — scaled to fit the node

Vim spatial navigation

Navigate the canvas without touching the mouse:

Monaco text editor inside nodes

Double-click any text node to edit it inline — full Monaco editor with vim keybindings, markdown syntax highlighting, and VS Code theme integration.

Vault integration

Point Skena at a folder of .md files (a vault). Use Ctrl+N to fuzzy-search and add any entry as a node. The vault is watched for changes — nodes update automatically.

Canvas portals

Select nodes → right-click → Move to sub-canvas — selected nodes move to a new .canvas file and a portal node replaces them. Click the portal to open the linked canvas in a new tab.

Zoom-adaptive rendering (LOD)

MCP server

On activation, Skena auto-deploys a local MCP server to .vscode/skena-mcp.js so Claude Code and other AI tools can read and write canvas nodes programmatically.

VS Code Remote SSH compatible

All processing runs in the extension host on the remote machine. No local toolchain needed.

Getting Started

1. Install the extension from the marketplace
2. Create or open any .canvas file — Skena opens it automatically as a visual canvas
3. (Optional) Configure your vault in .vscode/settings.json:

{
  "skena.vaults": [
    { "name": "v1", "path": "~/vault" }
  ],
  "skena.vaultDirectories": ["notes", "knowledge", "logs"]
}

4. Press Ctrl+N on the canvas to fuzzy-search your vault and add nodes

Configuration

{
  // Named vaults — supports multiple. URI: vault://<name>/path/to/file.md
  "skena.vaults": [
    { "name": "v1", "path": "~/vault" }
  ],

  // Directories to scan in each vault
  "skena.vaultDirectories": ["alpha", "knowledge", "logs", "inbox"],

  // Default node dimensions for newly created nodes
  "skena.nodeWidth": 400,
  "skena.nodeHeight": 250,

  // Auto-save debounce (ms)
  "skena.autoSaveDelay": 500,

  // Show source cells alongside notebook outputs
  "skena.notebook": {
    "showSourceCells": false
  }
}

Canvas Format

Skena reads and writes standard JSON Canvas 1.0 — the same format used by Obsidian. Files are plain JSON, version-control friendly, and diffable.

Supported node types: file, text, group, link, cell (standalone output), chat (AI terminal, UI only), portal (link to another canvas)

Extension node types (cell, chat, portal) are silently ignored by Obsidian, so files remain fully interoperable.

Obsidian Compatibility

.canvas files created in Skena open correctly in Obsidian and vice versa. Vault .md files use standard YAML frontmatter — readable by Obsidian, Foam, and any markdown tool.

License

MIT — github.com/dmarienko/skena