🤖 QuickAI for PowerToys Run

February 12, 2026 · View on GitHub

🤖 QuickAI for PowerToys Run

⚡ AI-Powered Assistance Right at Your Fingertips ⚡

Never leave your workflow to get AI help again.
Access multiple AI providers instantly with Alt+Space → ai explain quantum computing → Done! ✨

Maintenance

❌ Before	✅ With QuickAI
Open browser → Login to AI → Type question → Copy	`Alt+Space` → `ai speed of light` → Copy
30+ seconds, context switching	2 seconds, zero interruption

🧭 Table of Contents

Overview
Recent Changelog
Features
Supported AI Providers
Demo
Quick Start
Installation
Configuration
Usage Examples
Advanced Features
Building from Source
Development & Quality
Project Structure
FAQ
Troubleshooting
Security & Privacy
Tech Stack
Contributing
Screenshots
Support the Project
License
Acknowledgements

QuickAI brings the power of multiple AI providers directly into PowerToys Run. Get instant, streaming responses from leading AI models without leaving your workflow. Launch PowerToys Run with Alt+Space, type ai (or your custom keyword), and ask anything.

Plugin ID: 420129A62ECA49848C5C7CA229BFD22C
Action Keyword: ai or change to quickai
Platform: Windows 10/11 (x64, ARM64)
Tech: C#/.NET 9.0, WPF, PowerToys Run API

📋 Recent Changelog

⚡ v1.2.1 - Quick Commands & Clipboard Text Auto-Read

Quick Commands System:

Type ai / to see all available commands instantly
Built-in commands: /read, /explain, /summarize, /translate, /fix
Clipboard Text Auto-Read: Commands automatically grab text from clipboard when no extra text typed
- ai /translate → translates clipboard text
- ai /summarize → summarizes clipboard content
Custom Quick Commands: Add your own commands in settings
- Format: /name=prompt; /name2=prompt2
- Example: /review=Review this code for best practices; /eli5=Explain like I'm 5
UI Fix: Custom Quick Commands setting now visible (fixed Checkbox type issue)

💬 v1.2.0 - System Prompt, Clipboard Image & Quick Commands

📸 Clipboard Image Attachment

Attach screenshots directly to your queries — AI reads and analyzes image content!

PowerToys Run with clipboard image indicator

📎 Clipboard image indicator shows when image attachment is active
QuickAI reading text from screenshot

QuickAI Response window — send "read text from image" with a clipboard screenshot

New Features:

💬 Custom System Prompt: Persistent instructions sent before every query (e.g. "You are a concise assistant")
- OpenAI: role: system · Google: systemInstruction · Cohere: preamble
📸 Clipboard Image Attachment: Send clipboard images with queries for vision models
- OpenAI: image_url · Google: inlineData · UI shows 📎 when active
⚡ Quick Commands: Type / to see built-in commands — no need to type full prompts!
⚙️ Custom Commands: Add your own /commands in settings

⚡ Quick Commands

Command	Description
`/read`	📸 Read text from clipboard image
`/explain`	💻 Explain code step by step
`/summarize`	📝 Summarize content concisely
`/translate`	🌍 Translate to English
`/fix`	🐛 Find and fix bugs in code

Add your own in settings (Custom Quick Commands):

/review=Review this code for best practices; /eli5=Explain like I'm 5

Thanks to @ELGUAPOLIFE for the feature request (#13)!

🎨 v1.1.1 - UI Overhaul & Enhanced Experience

Major UI Improvements:

🪟 New Dedicated Results Window: Replaced static MessageBox with a beautiful, full-featured WPF window
📝 Markdown Rendering: Proper support for bold, italic, code blocks, and inline code
🎨 Modern Typography: Inter font for UI and JetBrains Mono for code (matching Cursor editor style)
🌓 Dynamic Theme Support: Automatic light/dark mode with real-time theme updates when system theme changes
🔄 Real-time Streaming: Token-by-token updates directly in the results window
📋 Enhanced UX: Copy button, word wrap toggle, character counter, and keyboard shortcuts (Ctrl+C, Esc)
🧹 Clean Interface: Query box automatically clears when opening results window
📜 Auto-scroll: Automatically scrolls to bottom as new content streams in

Thanks to @WuJunkai2004 (Wu Junkai) for the ideas and contributions!

🎉 v1.1.0 - Local AI with Ollama

Major Feature:

🏠 Local AI Support: Run AI models completely offline on your own machine with zero API costs
🔒 Full Privacy: All processing happens locally - no data leaves your computer
💰 Zero Cost: Unlimited queries with no API fees
🚀 Fast: No network latency, instant responses
🔧 Flexible: Use any Ollama model - llama3.2, mistral, qwen2.5, deepseek-r1, etc.
⚙️ Customizable: Configure custom host URL for remote Ollama instances

Thanks to @fuzhiyin-7 for suggesting this feature!

⚡ Features That Matter

⚡ Streaming Responses _{Real-time token-by-token display as you type}	🌐 Multi-Provider _{Groq, Together, Fireworks, OpenRouter, Cohere, Google, Ollama}	🔑 Dual API Keys _{Primary key with automatic secondary key fallback}
⚙️ Highly Configurable _{Customize model, temperature, max tokens, and timeout}	📋 One-Click Copy _{Press Enter to copy responses to clipboard}	💬 Custom System Prompt _{Persistent instructions before every query}
📸 Clipboard Images _{Attach screenshots for vision AI models}	🪟 Results Window _{Dedicated window with markdown rendering}	🏠 Local AI _{Run models offline with Ollama}

🤖 Supported AI Providers

Provider	Endpoint	Default Model	API Schema
Groq	`https://api.groq.com/openai/v1/chat/completions`	`llama-3.1-8b-instant`	OpenAI
Together	`https://api.together.xyz/v1/chat/completions`	`llama-3.1-8b-instant`	OpenAI
Fireworks	`https://api.fireworks.ai/inference/v1/chat/completions`	`llama-3.1-8b-instant`	OpenAI
OpenRouter	`https://openrouter.ai/api/v1/chat/completions`	`llama-3.1-8b-instant`	OpenAI
Cohere	`https://api.cohere.com/v1/chat`	`command`	Cohere
Google	`https://generativelanguage.googleapis.com/v1beta`	`gemini-1.5-flash`	Google
Ollama 🆕	`http://localhost:11434/v1/chat/completions`	`llama3.2`	OpenAI

🎉 New in v1.1.0: Local AI with Ollama support! Run models on your own machine with zero API costs. Thanks to @fuzhiyin-7 for the idea!

🎨 New in v1.1.1: Dedicated Results Window with markdown rendering, modern typography (Inter & JetBrains Mono), dynamic theme support, and enhanced UX! Thanks to @WuJunkai2004 for the ideas!

💬 New in v1.2.0: Custom System Prompt & Clipboard Image Attachment! Configure persistent instructions and attach screenshots for vision models. Thanks to @ELGUAPOLIFE for the feature request!

📸 Demo Gallery

AI Assistant in Action

Streaming Response

Plugin Settings

Multiple Use Cases

New Results Window (v1.1.1)

New Results Window with Markdown Support

Clipboard Image & Quick Commands (v1.2.0)

_{📎 Clipboard image indicator in PowerToys Run} QuickAI reading text from screenshot

_{QuickAI Response window with clipboard image query}

🏃‍♂️ Get Started in 60 Seconds

1️⃣ Download & Extract (15 seconds)

# Choose your architecture
wget https://github.com/ruslanlap/PowerToysRun-QuickAi/releases/download/v1.2.1/QuickAi-1.2.1-x64.zip
# Extract to PowerToys plugin directory

2️⃣ Auto-Install Location (30 seconds)

%LOCALAPPDATA%\Microsoft\PowerToys\PowerToys Run\Plugins\QuickAi

3️⃣ Restart PowerToys (15 seconds)

Right-click PowerToys tray icon → Exit → Restart from Start Menu

4️⃣ Configure API Key 🎯

Open PowerToys Settings → PowerToys Run → Plugins → QuickAI
Enter your API key for the preferred provider
Hit Alt+Space → Type ai what is quantum computing? → Marvel at the magic! ✨

Pro Tip: Change the keyword from ai to q or ask in PowerToys settings for even faster access!

💡 Power User Commands

# 🧠 Complex Explanations
ai explain machine learning in simple terms
ai how does blockchain work?
ai what is the difference between REST and GraphQL?

# 💻 Code Generation & Help
ai write a python function to sort a list
ai debug this javascript code: [paste code]
ai convert this sql to mongodb query

# 📝 Text Processing
ai summarize this text: [paste text]
ai improve the grammar of this sentence
ai translate "hello world" to spanish

# 🎯 Creative Tasks
ai write a poem about programming
ai brainstorm ideas for a mobile app
ai create a product description for headphones

# ⚡ Quick Commands (NEW! Type / to see all)
ai /read          # 📸 Read text from clipboard screenshot
ai /explain       # 💻 Explain code from clipboard image
ai /summarize     # 📝 Summarize content
ai /translate     # 🌍 Translate to English
ai /fix           # 🐛 Find and fix bugs in code

# 📸 Vision & Screenshot Tasks
ai what does this error message mean? 📎 (with clipboard screenshot)
ai explain this code snippet 📎 (with clipboard screenshot)
ai read the text from this image 📎 (with clipboard screenshot)

# 💬 System Prompt Examples
# Set system prompt to "You are a concise assistant" then:
ai explain quantum computing
# Set system prompt to "You are a helpful coding tutor" then:
ai debug this python code

# ⚙️ Custom Commands (add your own in settings!)
# /review=Review this code for best practices
# /eli5=Explain like I'm 5
ai /review        # Uses your custom prompt
ai /eli5          # Uses your custom prompt

🎨 See All Available Features

Streaming Responses: Watch answers appear in real-time (updates every 3 tokens or 150ms)
Context Menu: Right-click results for:
- Show full response (Enter)
- Copy response (Ctrl+C)
- Restart query (Ctrl+R)
Multi-language Support: Works with any language
Code Highlighting: Syntax-aware responses for code snippets
History: Previous responses cached for quick reference
Custom Models: Use any model from supported providers

🔧 Advanced Configuration

🎯 Customize Your Experience

Change Trigger Keyword

PowerToys Settings → PowerToys Run → Plugins → QuickAI → Change ai to your preference

Optimize Performance

Temperature: Adjust creativity (0.0 = factual, 2.0 = creative, default: 0.2)
Max Tokens: Control response length (16-4096, default: 128)
Request Timeout: Configure timeout in seconds (3-30, default: 8)
Model Selection: Choose specific models per provider
Dual API Keys: Configure backup keys for reliability

🆕 New Advanced Features

System Prompt: Configure persistent instructions sent before every query (e.g., "You are a concise assistant")
- Works across all providers with native API support
- Perfect for setting consistent AI personality or response style
Clipboard Image Attachment: Toggle to attach clipboard images with queries
- Requires vision-capable models (GPT-4V, Gemini Pro Vision, etc.)
- Shows 📎 indicator when enabled
- Great for code analysis, error messages, charts, and visual content
Quick Commands: Type / to see built-in commands (/read, /explain, /summarize, /translate, /fix)
- Add your own custom commands in settings
- Format: /name=prompt; /name2=prompt2
- Append extra text: ai /translate to French sends "Translate the following to English: to French"

Provider-Specific Settings

Groq: Best for speed with Llama models
Together: Wide variety of open-source models
Fireworks: Fast inference with custom models
OpenRouter: Access to multiple providers via one key
Cohere: Enterprise-grade models with different API
Google: Google Gemini models (API key passed as URL parameter)
Ollama 🆕: Run models locally on your machine - zero API costs, complete privacy, customizable host URL

⚠️ Important: Model Name Configuration

Critical: Always copy the exact model name from your provider's documentation or dashboard. Model names vary by provider and must match exactly:

Examples of correct model names:

Groq: llama-3.1-8b-instant, mixtral-8x7b-32768
Together: meta-llama/Llama-3-8b-chat-hf, mistralai/Mixtral-8x7B-Instruct-v0.1
Fireworks: accounts/fireworks/models/llama-v3p1-8b-instruct ⚠️ (must include full path)
OpenRouter: meta-llama/llama-3.1-8b-instruct, anthropic/claude-3-haiku
Cohere: command, command-light, command-nightly
Google: gemini-1.5-flash, gemini-1.5-pro, gemini-pro
Ollama 🆕: llama3.2, mistral, qwen2.5, codellama, deepseek-r1 (any locally pulled model)

_{Example: Fireworks requires full model path like accounts/fireworks/models/llama-v3p1-8b-instruct}

Where to find correct model names:

Groq: console.groq.com/docs/models
Together: docs.together.ai/docs/inference-models
Fireworks: docs.fireworks.ai/models - Copy from "Model ID" field
OpenRouter: openrouter.ai/models
Cohere: docs.cohere.com/docs/models
Google: ai.google.dev/models - Google Gemini models
Ollama 🆕: ollama.com/library - Run ollama list to see locally pulled models

🔑 API Key Acquisition

How to Get Your API Keys

Groq

Visit https://console.groq.com
Sign up or log in
Navigate to API Keys section
Create a new API key
Copy and paste into QuickAI settings

Together

Visit https://api.together.xyz
Create an account
Go to Settings → API Keys
Generate a new key
Copy and paste into QuickAI settings

Fireworks

Visit https://fireworks.ai
Sign up for an account
Navigate to API Keys
Create a new key
Copy and paste into QuickAI settings

OpenRouter

Visit https://openrouter.ai
Sign up or log in
Go to Keys section
Generate an API key
Copy and paste into QuickAI settings

Cohere

Visit https://cohere.com
Create an account
Navigate to API Keys
Create a new key
Copy and paste into QuickAI settings

Google

Visit https://ai.google.dev
Sign up or log in with your Google account
Navigate to Get API Key section
Create a new API key
Copy and paste into QuickAI settings

Ollama (Local - No API Key Required!) 🆕

Install Ollama from https://ollama.com
Run ollama serve to start the local server
Pull a model: ollama pull llama3.2 (or any model)
Select "Ollama" provider in QuickAI settings
Set model name to your pulled model (e.g., llama3.2, mistral, qwen2.5)
Optionally configure custom host URL (default: http://localhost:11434)

💡 Tip: Most cloud providers offer free tiers perfect for testing QuickAI! With Ollama, run unlimited queries locally for free!

🔧 Technical Details

Architecture & Implementation

Architecture

Target Framework: .NET 9.0 Windows (10.0.22621.0)
Plugin Type: PowerToys Run Plugin (DynamicLoading: false)
HTTP Client: Static singleton with configurable timeout (default: 8 seconds, range: 3-30)
Streaming: Server-Sent Events (SSE) parsing with real-time UI updates
JSON: System.Text.Json for serialization
UI: WPF with theme-aware design

Implementation Highlights

Implements IPlugin for PowerToys Run integration
Implements ISettingProvider for GUI configuration via AdditionalOptions
Implements IContextMenu for right-click actions (show full response, copy, restart)
Implements IDisposable for proper resource cleanup
Async HTTP requests with Task.Run() for non-blocking queries
Real-time result updates during streaming (batched every 3 tokens or 150ms for performance)
Automatic retry with secondary API key on failure
Thread-safe response building with StringBuilder
Supports multiple API schemas (OpenAI, Cohere, Google Gemini)

Error Handling

Validates API key presence before requests
Handles network failures gracefully
Shows user-friendly error messages
Configurable timeout protection (default: 8 seconds, range: 3-30)
JSON parsing error recovery
Automatic fallback to secondary API key

Project Structure

PowerToysRun-QuickAi/
├── QuickAi/
│   ├── Community.PowerToys.Run.Plugin.QuickAi/
│   │   ├── Main.cs                    # Core plugin implementation (1275 lines)
│   │   ├── plugin.json                # Plugin manifest
│   │   ├── *.csproj                   # Project file
│   │   └── Images/
│   │       ├── ai.dark.png           # Dark theme icon
│   │       └── ai.light.png          # Light theme icon
│   └── QuickAi.sln                   # Solution file
├── assets/                            # README images and demos
├── .github/workflows/                 # CI/CD automation
└── README.md                          # This file

🏗️ For Developers

🚀 Quick Build

git clone https://github.com/ruslanlap/PowerToysRun-QuickAi.git
cd PowerToysRun-QuickAi
./build-and-zip.sh  # Creates distribution-ready packages

🧪 Tech Stack

Runtime: .NET 9.0 with Windows 10/11 target
UI: WPF with theme-aware design
HTTP: HttpClient with streaming support
Architecture: Service-oriented with clean separation
Serialization: System.Text.Json for API responses

🎯 Project Highlights

Clean Code: Well-documented, testable architecture
Multi-Platform: x64 + ARM64 builds
Streaming: Real-time response handling
Plugin Template: Great starting point for other PowerToys plugins
Provider Abstraction: Easy to add new AI providers

Want to contribute?

🌟 Community & Support

Show Some Love ❤️

If QuickAI boosts your productivity, consider starring the repo and sharing with colleagues!

Support Development ☕

Join the Community

🆘 Troubleshooting

Plugin not showing up?

✅ Check path: %LOCALAPPDATA%\Microsoft\PowerToys\PowerToys Run\Plugins\QuickAi
✅ Verify files: Ensure plugin.json exists in the folder
✅ Restart: Completely exit and restart PowerToys
✅ Permissions: Run PowerToys as administrator once

API requests failing?

✅ API Key: Verify your API key is correct and active
✅ Internet: Check connection for API requests
✅ Provider: Ensure selected provider is available
✅ Model: Verify model name is supported by provider

Streaming not working?

✅ Network: Check if streaming is blocked by firewall
✅ Timeout: Increase timeout in plugin settings
✅ Provider: Some providers may have streaming limitations

"Show full response" only shows partial text?

From issue #3: this is expected when the AI response is truncated by settings, not a UI bug. The provider stops early if token limits or timeouts are too low, so the modal only shows the first words.

What to do

✅ Increase Max Tokens: Try 2048–4096 to allow full answers.
✅ Raise Request Timeout: Low timeouts can cut responses.
✅ Provider limits: Some providers enforce their own caps.
ℹ️ Temperature: Only affects creativity, not length.

If you still need the full text, press Ctrl+C in the result list to copy the complete response.

❓ FAQ

How do I change the AI provider?

PowerToys Settings → PowerToys Run → Plugins → QuickAI → Select provider from dropdown

Can I use multiple API keys?

Yes! Configure both primary and secondary API keys. The plugin will automatically fallback if the primary fails.

What models are supported?

Any model supported by your chosen provider. Popular choices include Llama 3.1, Mixtral, and provider-specific models.

Is my data private?

Your queries are sent to the chosen AI provider's API. Check each provider's privacy policy for details on data handling. **NEW**: With Ollama, everything stays on your machine - complete privacy!

Can I use it offline?

Cloud providers (Groq, Together, etc.) require internet. **NEW**: With Ollama, you can run AI completely offline on your local machine!

How do I use Quick Commands?

Type ai / to see all available commands. Built-in commands include /read, /explain, /summarize, /translate, /fix. You can also add your own custom commands in PowerToys Settings → QuickAI → Custom Quick Commands using the format: /name=prompt; /name2=prompt2

How do I use Ollama for local AI?

1. Install Ollama from [ollama.com](https://ollama.com)
2. Run `ollama serve` in terminal
3. Pull a model: `ollama pull llama3.2`
4. Select "Ollama" in QuickAI settings
5. Set model name to your pulled model
6. Enjoy unlimited local AI queries with zero API costs!

🔒 Privacy & Security

✅ Local Plugin: All plugin functionality runs locally
✅ No Tracking: Zero analytics or telemetry from the plugin
✅ Open Source: Full code transparency
✅ MIT License: Free for any use

⚠️ API Key Storage

Important Security Notice: API keys are currently stored in plain text in PowerToys settings JSON file:

%LOCALAPPDATA%\Microsoft\PowerToys\PowerToys Run\settings.json

Security Recommendations:

✅ Use API keys with limited permissions or quotas
✅ Do not use production or billing-critical keys
✅ Consider rotating keys periodically
✅ Avoid sharing your settings.json file
✅ Use provider-specific key restrictions (IP allowlists, rate limits)

Future Enhancement: Windows DPAPI encryption for API keys is planned for a future release.

Data Privacy

Your queries are sent to the chosen AI provider's API
QuickAI plugin itself does not collect or store any query data
Check each provider's privacy policy for their data handling practices
Most providers use queries for model improvement (check provider settings to opt-out)
🆕 Ollama: Complete privacy - all processing happens locally on your machine, no data leaves your computer

📄 License

Released under the MIT License. Free for personal and commercial use.

🙏 Acknowledgements

Powered by amazing AI providers:

Groq - Fast inference platform
Together AI - Open-source model platform
Fireworks AI - AI model inference
OpenRouter - Unified AI API access
Cohere - Enterprise AI models
Google AI - Google Gemini models
Ollama - Local AI model runner

Built with:

Microsoft PowerToys - The extensible Windows productivity toolkit
.NET - Cross-platform development platform

Special Thanks:

@fuzhiyin-7 - For suggesting Ollama integration, bringing local AI support to QuickAI! 🎉
@WuJunkai2004 (Wu Junkai) - For ideas and contributions that inspired the major UI overhaul with dedicated ResultsWindow, markdown rendering, and improved user experience in v1.1.1! 🎨✨
@ELGUAPOLIFE - For the feature request that brought custom system prompts and clipboard image attachment to QuickAI in v1.2.0! 💬📸

🚀 Ready to supercharge your workflow with AI?

_{Made with ❤️ by @ruslanlap}