In my previous post, I introduced PeepNote, a desktop app for keeping useful LLM answers and finding them again while working.
This time, I added Magic Pen to make asking the next question easier.
While reading documentation or browsing a website, I often find something I want an AI to explain. Usually, that means copying text or taking a screenshot, opening a chat, and pasting everything there.
I wanted to make that process shorter.
Draw around what you want to understand
Press Alt+S and draw around part of your screen.
Magic Pen sends the selected area to your configured AI and opens a movable window with a streaming response.
You can use it to understand a diagram, explain unfamiliar terminology, or organize information from a document.
For example, while reading about RAG, you could select an architecture diagram and then ask:
How is RAG different from a regular chatbot? Compare them in a simple table.
Keep the conversation going
The Reply button lets you continue in two ways:
- Text: type a follow-up question.
- Magic Pen: select another area and include it in the same conversation.
When the explanation is worth keeping, choose a workspace and click Save. The conversation becomes a Markdown note in PeepNote.
Choose your AI
Magic Pen supports your configured AI API or an installed, signed-in CLI tool: Codex, Claude Code, Gemini CLI, or Qwen Code.
You can choose the mode and model in Settings. Selected screen content is sent to the AI service you use, while saved notes stay on your computer.
PeepNote is free, open source, and available for Windows and Linux.
What would you use a screen-selection AI tool for? I’d love to hear your feedback and ideas for making it more useful.
Top comments (0)