PromptIQ

PromptIQ is a VS Code extension for developers who want to craft stronger prompts before sending them to Copilot, Claude, Cursor, or any other AI assistant. It combines prompt scoring, optimization, token and cost estimation, local history, and quick handoff actions into a single workflow.

Overview

PromptIQ helps you:

• score a prompt for clarity, specificity, actionability, and scope
• optimize a prompt into a clearer, more structured version
• estimate token usage and approximate cost before sending it
• keep a local history of optimized prompts for reuse
• copy the result or send it directly to Copilot Chat from inside VS Code

Current features

Implemented

• Sidebar-based prompt composer and analyzer UI
• Prompt scoring with actionable improvement suggestions
• Prompt optimization with structured output and rationale
• Token and cost estimation helpers
• Local prompt history persistence
• Developer logging and debug support
• Standalone webview development workflow for UI iteration
• Automated tests for the prompt engine logic

Planned / roadmap

• Chat participant support inside Copilot Chat
• MCP server integration for Claude Code and Cursor workflows
• Multi-provider model abstraction for OpenAI, Anthropic, Google, and VS Code LM
• Shared prompt libraries and richer analytics for team use

Project structure

• src/extension/ — extension activation, commands, and webview bridge
• src/services/ — scoring, optimization, history, logging, and supporting logic
• src/shared/ — shared message contracts between the extension host and webview
• src/webview/ — React-based sidebar UI and rendering entry points
• src/test/ — automated tests for the core engine behavior
• webview-dev/ — standalone dev server for the webview outside VS Code
• media/ — icons and static assets for the extension

Installation

1. Install dependencies:
   • npm install
2. Build the extension:
   • npm run compile
3. Launch the extension from VS Code using the Run Extension debug configuration

Development workflow

• npm run compile — bundle the extension and webview, then run TypeScript compilation
• npm run dev:webview — start the standalone webview dev server for local UI work
• npm run build:webview — build the webview bundle only
• npm run test — compile and run the test suite
• npm run vscode:prepublish — prepare the extension for packaging/publishing

Extension commands

The extension contributes the following commands:

• PromptIQ: Open Sidebar
• PromptIQ: Optimize Prompt
• PromptIQ: Copy Optimized Prompt
• PromptIQ: Send to Copilot Chat
• PromptIQ: Send to PromptIQ
• PromptIQ: Toggle Developer Mode
• PromptIQ: Open Developer Log

Configuration

PromptIQ supports the following settings:

• promptiq.defaultModel — default model name used for recommendations
• promptiq.enableHistory — enable or disable local prompt history persistence
• promptiq.developerMode — enable developer logging for debugging

Architecture at a glance

The extension is split into three main layers:

• Extension host — registers commands, manages state, and communicates with the webview
• Services — contain the prompt scoring, optimization, token, cost, and history logic
• Webview UI — renders the sidebar experience in React inside a VS Code webview

This separation keeps the core prompt logic testable and independent from the VS Code UI layer.

Testing

The project uses Node’s built-in test runner against the compiled TypeScript output. The current suite validates core scoring, optimization, token estimation, and cost logic.

Roadmap

PromptIQ is currently positioned as a Phase 1-style MVP centered on prompt intelligence and handoff flows. The next logical steps are:

1. expand the experience into a richer Copilot Chat participant integration
2. add MCP-based workflows for other AI tools
3. introduce model-provider abstraction and comparison features
4. grow toward team templates, analytics, and shared prompt libraries