AI Context Stacker

Prepare file context and directory maps for models like ChatGPT, Claude, or Gemini without leaving VS Code. Drag files into organized tracks to copy complete project context in one action.

The Problem

When working with LLMs, you typically need to:

1. Open a file and copy its contents
2. Switch to your browser
3. Paste it into the chat
4. Return to VS Code
5. Repeat for each additional file

This gets tedious fast—especially with larger codebases or multi-file tasks.

Quick Start

1. Install the extension from the VS Code Marketplace.
2. Open the AI Context Stacker view in the Activity Bar.
3. Stage Files by dragging them into the Staged Files panel.
   • Or right-click any file in the Explorer → Add to AI Context Stack.
4. Quick Add files from anywhere using Ctrl+Alt+A (Cmd+Alt+A on Mac).
5. Preview your context with Ctrl+Shift+V (Cmd+Shift+V on Mac, except in Markdown files).
6. Copy Stack by clicking the copy icon or pressing Ctrl+Shift+C (Cmd+Shift+C on Mac) when focused on the Staged Files view.
7. Paste into your LLM.

Tip: Press F1 or Ctrl+Shift+P (Cmd+Shift+P on Mac) and type "AI Context Stacker" to discover all available commands.

Key Features

File Staging

Drag files or folders into the staging area. Right-click any file in the Explorer to add it. The extension shows what's staged and lets you copy everything in one go.

Live Token Updates: Token counts refresh automatically as you edit (400ms debounce). Files over 1MB use optimized statistical estimation to keep the editor responsive during rapid typing.

Startup Behavior: The sidebar appears immediately when VS Code opens. Token analysis runs in the background during a brief warmup period, showing a loading spinner and "Calculating..." in the Status Bar until complete.

Manual Refresh: Use the dedicated Refresh Stack command (found in the ... menu or press Ctrl+Alt+U / Cmd+Alt+U) to force a re-scan of the filesystem and re-calculate all token counts.

Context Tracks

Create separate tracks for different tasks (e.g. Bug Fix #123, Refactor Auth). Each track maintains its own list of staged files.

• Reorder tracks by dragging or using Alt+↑ / Alt+↓
• Rename tracks inline with F2
• Quickly switch between tracks with Ctrl+Alt+S (Cmd+Alt+S on Mac)

Note: The extension always maintains at least one active track. The last remaining track cannot be deleted.

Multi-root Workspace Support

Designed for professional workflows, AI Context Stacker fully supports Multi-root Workspaces.

Staged files are grouped by their project folder name so the LLM understands which project each file belongs to—even when multiple projects contain identical paths like src/index.ts.

Works seamlessly across GitHub Codespaces, WSL2, and SSH Remote sessions with optimized clipboard and drag-and-drop handling for remote environments.

Token Warnings

Files are color-coded based on estimated token count:

• Amber — over 5,000 tokens (configurable via largeFileThreshold)
• Red — over 10,000 tokens (2x the threshold)
• Pinned files retain their pin icon while inheriting warning colors

This helps you spot large files before hitting model limits.

Pinning and Filtering

• Pin files to protect them from Clear Stack
• Toggle Show Pinned Files Only to filter the view
• Copy commands respect the active filter
• Use Space to quickly toggle pin on selected files

Pinned files persist across clears and workflow resets.

Context Map

Optionally include an ASCII directory tree in the copied output to help the LLM understand project structure.

# Context Map
├── components
│   ├── Header.tsx
│   ├── Sidebar.tsx
│   └── Footer.tsx
├── utils
│   ├── api.ts
│   └── helpers.ts
└── README.md

Selection-Aware Copying

Copy commands adapt to your current selection:

• Files selected → Copies only the selected files
• Nothing selected → Copies the entire visible stack (respects active filters)

This makes both targeted and bulk operations intuitive without separate commands.

Navigation

• Right-click files in the Explorer to add them or reveal them in the stack
• Right-click staged files to reveal them in your system file manager
• Use Reveal in AI Stack to locate and highlight any file within your staged context

Performance

The extension handles large stacks without blocking VS Code. The sidebar renders immediately on startup while token counting happens in the background. Folder trees with hundreds of files build quickly, and long operations show progress using VS Code's native progress bar.

Commands

Stack Operations

Output & Clipboard

View & Filtering

Track Management

Keyboard Shortcuts

Global Shortcuts

Staged Files View (when focused)

Context Tracks View (when focused)

Settings

All configuration is managed natively via VS Code Settings.

Tips

• Sidebar Placement: Keeping the extension in the left Activity Bar improves drag-and-drop from the Explorer.
• Model Limits: Adjust largeFileThreshold to match your model’s context window.
• Fast Iteration: Use Copy and Clear Stack (Ctrl+X / Cmd+X) to reset quickly between prompts.
• Selective Copy: Copy respects selection—select files to copy only those, or copy everything when nothing is selected.
• Bulk Actions: Use Ctrl/Cmd+A then Space to pin or unpin multiple files at once.

Known Limitations

• Clipboard Size: Output is capped at 100MB to protect VS Code stability.
• Large Files: Files over 5MB are excluded from context.
• Token Estimates: Counts are approximate and may differ slightly from model-specific tokenizers.
• Binary Files: Binary assets are automatically skipped.
• Large Folders: Adding folders with more than 200 files requires confirmation.
• Web VS Code: Clipboard access in browser-based VS Code requires a secure HTTPS context.

Support

• Issues: GitHub Issues
• Changelog: CHANGELOG.md

License

This project is licensed under the MIT License.