A lightweight desktop app for Windows and Linux that lets you press a hotkey, drag a box around any region of your screen, and get an AI-generated explanation right beside it—no switching apps, opening a chat tab, or manually uploading screenshots. Built with Rust and Tauri v2, it uses a native screenshot solution and streams answers from OpenAI models (Fast mode for speed, Best mode for quality). It supports multiple selection modes (Rectangle, Freeform Lasso, Multi-Region), follow-up questions, context chaining (pin a selection as context for the next query), markdown/code rendering, and local chat history. macOS is listed as “coming soon” but not yet shipped; Linux support is Wayland-only (GNOME 45+ or KDE Plasma 6), no X11. Pricing: Free tier with 3 captures/day and 10/week on Fast model (no card required); Plus at $10/month or $99/year with 2,500 credits/month, Best model access, and up to 10 follow-ups per capture. Plus is sold in the United States only, and payments are non-refundable.
Pros & Cons:
Pros:
Eliminates the screenshot → switch app → upload → switch back loop that makes screen-based AI queries tedious; native Rust/Tauri build with built-in screen capture is fast and stable, with Fast mode typically starting to answer in 1–2 seconds; context chaining lets you pin one selection and reference it in a later query (e.g., pin a database schema, then ask about a slow query); multiple selection modes including Freeform Lasso and Multi-Region support irregular shapes and side-by-side comparison; local chat history and markdown/code rendering make it practical for debugging and research; free tier is a real ongoing plan (not an expiring trial), and no card is required.
Cons: macOS is not yet available—the vendor still lists it as “coming soon,” so Mac users cannot use it today; Linux support is Wayland-only (no X11), which excludes many Linux setups; Plus is sold in the United States only, and payments are non-refundable (cancelling stops future charges but does not return the current period); requires sending selected screen regions to OpenAI’s API, which some users may find unacceptable for sensitive content; credit maths can be confusing—Fast costs 1 credit, Best costs 25 (or 20 in some sources), so Best usage burns through the 2,500 monthly credits quickly; no independent third-party reviews surfaced beyond directory listings and a tutorial, so long-term reliability is unverified.
Who is it for?:Windows and Linux users who repeatedly need to understand something already on screen (code, charts, error dialogs, UI elements, PDF paragraphs) without rebuilding context in a separate chatbot; developers debugging stack traces or asking about code snippets; designers reviewing UI or accessibility issues; students and researchers extracting information from dense documents; anyone who currently screenshots into ChatGPT several times a day and wants that loop collapsed.