Prompt Queue + Usage Monitor

A VS Code extension that lets you queue prompts for later delivery (ideal when you hit a rate limit) and monitor your AI token usage (Claude / OpenAI Codex) directly in the editor.

Features

A — Send Prompt After Rate-Limit

Queue any prompt now; the extension automatically creates a ready-to-use Markdown file in your workspace when the timer expires.

• Select text in any editor → PromptQueue: Queue Prompt (Send Later)
• Or paste from clipboard → PromptQueue: Queue From Clipboard
• Or take the whole current file → PromptQueue: Queue From Current Editor
• Auto-detects "try again in X hours" in clipboard text and pre-fills the delay
• File is created in .prompt-queue/YYYYMMDD_HHMM_<id>.md with a clean header
• Survives VS Code restarts (queue is persisted in global state)

B — Token Usage Monitor

View your last-5-hour and last-7-day token consumption via a panel in the Activity Bar.

• OpenAI usage via the organization Usage API
• Anthropic usage via the Admin API
• Local estimate (no keys needed) — counts queued prompt tokens as a rough proxy

Quick Start

1. Install and Open

# From source
npm install

Press F5 → a new VS Code Extension Development Host opens.

Click the clock icon in the Activity Bar to open the Prompt Queue panel.

2. Queue a Prompt

1. Write or select your prompt in any editor.
2. Open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P).
3. Run PromptQueue: Queue Prompt (Send Later).
4. Enter the delay in hours (default: 5).
5. Confirmation: "Prompt queued for 2024-06-15 20:00 [id: abc12345]".

When the time arrives, a file appears in .prompt-queue/ and a notification pops up with Open File / Reveal in Explorer actions.

3. Configure API Keys (optional, for accurate usage data)

Keys are stored in VS Code SecretStorage — never written to settings.json or disk in plain text.

OpenAI

1. Get an organization admin key from platform.openai.com/account/api-keys. (Standard project keys sk-proj-… cannot access the usage endpoint.)
2. Run: Usage: Set OpenAI API Key (Secret) from the Command Palette.
3. Paste your key → Enter.
4. Optionally set openai.orgId and openai.projectId in Settings.

Anthropic

1. Get an admin API key from console.anthropic.com.
2. Run: Usage: Set Anthropic Admin API Key (Secret).
3. Paste your key → Enter.
4. If your org requires it, set anthropic.orgId in Settings.

Verify

Run Usage: Refresh — the Usage Monitor panel updates with live data.

Commands

Configuration

All settings are under the promptQueue and usage namespaces (VS Code Settings UI or settings.json):

| Setting | Default | Description | |---|---|---| | promptQueue.defaultDelayHours | 5 | Default delivery delay (hours) | | promptQueue.outputDir | .prompt-queue | Workspace-relative output directory | | promptQueue.filenameTemplate | {timestamp}_{id}.md | Filename template ({timestamp} = YYYYMMDD_HHMM, {id} = 8-char hex) | | openai.orgId | | OpenAI Organization ID (optional) | | `openai.projectId` | | OpenAI Project ID (optional) | | anthropic.orgId | `` | Anthropic Org/Workspace ID (optional) | | usage.refreshIntervalMinutes | 10 | Auto-refresh interval; 0 = disabled |

How It Works

Queue persistence

Items are stored in vscode.ExtensionContext.globalState (a Memento backed by SQLite inside VS Code). They survive restarts and work across workspaces.

Delivery

Every 60 seconds (and on activation), the extension checks whether any queued item's notBefore timestamp is in the past. For each due item it:

1. Resolves the target workspace folder.
2. Creates <outputDir>/<filename> via workspace.fs (works on Remote / WSL / SSH).
3. Handles filename collisions (_2, _3, …).
4. Marks the item as processed.
5. Shows an actionable notification.

If VS Code is closed when the time arrives, the item is processed on the next activation.

Usage data

Data may lag by a few minutes (noted in the UI). The extension aggregates the last 5 hours and last 7 days of tokens.

Security

• No secrets in settings.json — all API keys use vscode.SecretStorage (OS keychain on macOS, Credential Manager on Windows, libsecret on Linux).
• No telemetry — this extension collects nothing.
• No hard-coded credentials.
• Network requests go only to api.openai.com and api.anthropic.com (when keys are configured).

Limitations (V1)

• Rate-limit detection is not automatic — the extension does not intercept Claude Code / Codex traffic. Use the clipboard/editor queue commands manually.
• The openai.adminApiKey must be an organization admin key, not a standard project key.
• The Anthropic usage endpoint path (/v1/organizations/{orgId}/usage) may change as the Admin API evolves. Check Anthropic docs if you get 404 errors.
• The usage monitor shows approximate data for queued prompts only (local estimate) when no API keys are set.

Development

# Install deps
npm install

# Compile (watch)
npm run watch

# Press F5 in VS Code to launch Extension Development Host

# Run unit tests (no VS Code instance needed for time/queue tests)
npm test

# Package as VSIX
npm run package

Project structure

src/
  extension.ts           ← Activation, command wiring
  queue/
    QueueStore.ts        ← Persistent queue (globalState)
    QueueProcessor.ts    ← Timer + file delivery
  usage/
    IUsageProvider.ts    ← Interface + types
    OpenAIUsageProvider.ts
    AnthropicUsageProvider.ts
    LocalEstimateProvider.ts
    UsageService.ts      ← Aggregation + caching
  ui/
    UsageViewProvider.ts ← TreeView for usage
    QueueViewProvider.ts ← TreeView for queue items
  util/
    time.ts              ← Pure time helpers
    crypto.ts            ← ID generation
    fs.ts                ← workspace.fs helpers
  test/
    runTest.ts
    suite/
      time.test.ts
      queue.test.ts

License

MIT