Visual Context Node Builder for AI

Stop blindly pasting files into ChatGPT. Take control of exactly what your AI sees.

An infinite-canvas node graph inside VS Code where you visually map your codebase, prune what's irrelevant, and export a perfectly structured context bundle — ready for any AI coding assistant.

Why This Exists

Every AI coding tool has the same problem: context. You either dump entire files and burn tokens, or you manually copy-paste snippets and lose the big picture. Neither works well.

This extension gives you a visual workspace to see your code relationships, choose exactly what goes in, and export a clean, structured prompt — with full control over every line.

Key Features

Node Graph Canvas

Drop files from the Explorer onto an infinite canvas. Each node shows the file's exports — functions, classes, types — and lets you toggle individual symbols on/off. Only include what matters.

One-Click Dependency Expansion

Hit Expand Deps on any file node and watch its import tree unfold across the canvas. Filter by category (source, styles, data) so you don't drown in noise.

Package Type Resolution

Need the AI to understand your dependencies? Add packages directly — the extension resolves .d.ts type definitions from node_modules and displays them inline.

Privacy Redaction

Mark any node as redacted. Its content becomes [REDACTED FOR PRIVACY] in the export — the file stays in the graph for structural context, but sensitive code never leaves your machine.

System Templates

Pre-built instruction sets for common tasks: Code Review, Refactor, Bug Fix, Explain Code, and New Feature. One click to set the AI's role and constraints.

Recipes

Save your canvas layouts as reusable recipes. Next time you need the same context setup, load it in one click instead of rebuilding from scratch.

Git-Aware Context Seeding

The fastest way to build context from what actually changed. Click Seed from Git and choose staged changes, unstaged changes, or any commit. The extension:

1. Parses the diff and identifies exactly which symbols (functions, types, classes) were added, modified, or deleted — not just which lines changed.
2. Scans your entire workspace for reverse dependencies — every file that imports or uses a changed symbol, with line numbers.
3. Populates the canvas automatically with color-coded nodes (green = added, amber = modified, red = deleted) and highlights changed symbols with orange dots.
4. Opens a Dependents Panel where you can selectively add consuming files, with dependency edges auto-drawn.
5. Includes a structured Git Diff Summary and Impact Analysis in the exported context.

Why this is better than pasting a diff:

	Raw git diff	Git Seed Context
Granularity	Line-level hunks with ~3 lines of context	Symbol-level: knows which functions changed
Blast radius	None — you guess what's affected	Full impact analysis: every file consuming each changed symbol
File content	Disconnected hunks	Complete file content with symbol selection
Effort	Manual copy-paste	One click

Use cases:

• Code review prep — seed from staged changes, get every changed file + every consumer on the canvas instantly
• Bug fix context — seed from the breaking commit, see the full function and all its call sites
• Refactoring impact — seed from unstaged changes, the reverse dependency scanner finds consumers in directories you'd never think to check
• PR descriptions — generate context from the seeded canvas and paste the structured summary straight into the PR

Live Token Estimation

A running token count updates as you add, remove, or modify nodes. Know exactly how much context you're sending before you send it.

Structured Export

Generate your context as Markdown or XML — optimized for how LLMs parse structured input. Copy to clipboard or save to file.

Getting Started

1. Install the extension
2. Open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P)
3. Run "Open Visual Context Builder"
4. Right-click any file in the Explorer and select "Add to Visual Context"
5. Expand dependencies, prune nodes, write your prompt
6. Click Generate Context and paste into your AI assistant of choice

AI-Powered Recipe Generation

Let your AI coding assistant build recipes for you. Instead of manually dragging files onto the canvas, ask your AI tool to analyze the codebase and generate a ready-to-use recipe. It searches your code, identifies relevant files and dependencies, and creates a .vcnb/recipes/<name>.json file you can open, edit, and export — all in one command.

The generated recipe includes:

• File nodes pointing to every relevant source file
• Sticky notes explaining key patterns and gotchas
• System instructions tailored to the task (review, debug, refactor, etc.)
• Dependency edges showing how files relate to each other

This works with any AI coding tool. Pick your setup below.

---

Claude Code

Claude Code gets a native /vcnb slash command for the smoothest experience.

Setup:

# From the extension repo
mkdir -p .claude/commands
cp skills/vcnb.md .claude/commands/

# Or download directly into any project
mkdir -p .claude/commands
curl -o .claude/commands/vcnb.md \
  https://raw.githubusercontent.com/pizan/visual-context-node-builder/main/skills/vcnb.md

Usage:

/vcnb the authentication flow
/vcnb all API route handlers and their middleware
/vcnb update auth-flow to include the middleware tests
/vcnb list

---

Cursor

Cursor loads instructions from .cursorrules at your project root.

Setup:

Append the contents of skills/vcnb-rules.md to your .cursorrules file:

# Download and append to your existing .cursorrules
curl -s https://raw.githubusercontent.com/pizan/visual-context-node-builder/main/skills/vcnb-rules.md >> .cursorrules

Usage:

Create a VCNB recipe for the authentication flow
Build a visual context map of the payment module
Update the auth-flow recipe to include tests

---

GitHub Copilot

Copilot reads project instructions from .github/copilot-instructions.md.

Setup:

mkdir -p .github
curl -s https://raw.githubusercontent.com/pizan/visual-context-node-builder/main/skills/vcnb-rules.md >> .github/copilot-instructions.md

Usage:

Create a VCNB recipe for the authentication flow
Map out the API routes and middleware as a visual context

---

Windsurf

Windsurf loads instructions from .windsurfrules.

Setup:

curl -s https://raw.githubusercontent.com/pizan/visual-context-node-builder/main/skills/vcnb-rules.md >> .windsurfrules

Usage:

Create a VCNB recipe for the authentication flow
Build a visual context for the data pipeline

---

Other AI Tools

For any AI tool that supports system-level instructions or custom prompts, use the universal rules file:

1. Download skills/vcnb-rules.md
2. Paste its contents into your tool's instructions/rules configuration
3. Ask the AI to "create a VCNB recipe for [your topic]"

The rules file contains the complete recipe JSON schema, layout guidelines, and quality checklist — everything the AI needs to generate valid recipes.

---

What happens after generation

1. Open the Visual Context Builder panel in VS Code
2. Go to the Recipe Library
3. Load your generated recipe
4. The canvas populates with all the files, notes, and connections
5. Edit, prune, or expand as needed
6. Export your context as Markdown or XML

Requirements

• VS Code 1.85+

Build from Source

npm install
npm run build

npm test        # run tests
npm run package # create .vsix

License

MIT