How to automatically send screenshots to GPT or Claude

If you've ever wished for an app that automatically screenshots a window and sends each shot to GPT or Claude so the AI just explains what's on your screen, that's exactly what Auto-Snapper AI does. You pick one window of your own study material, it captures on a timer, and every capture goes to OpenAI or Anthropic using your own API key. No copying, no pasting, no tab-switching.

It runs locally on macOS (Apple Silicon and Intel) and Windows. The app talks only to your chosen AI provider, with no middle-man server in between. Here's how to set it up end to end.

1. Pick the one window to capture

Select the exact window you want read from your own study materials, such as a PDF, a slide deck, your own practice-question set, a web page, an app, a terminal, or a remote session like RDP or Citrix where you study your own course or lab material that you have rights to use. A live preview confirms you grabbed the right one, and only that single window is ever captured. Nothing else on your screen is touched.

2. Bring your own GPT or Claude key

Auto-Snapper AI doesn't resell you tokens. You connect your own OpenAI or Anthropic API key and pay your provider directly at cost, with no per-question markup. You can switch models anytime, from a cheap and fast one for quick review to a frontier reasoning model for genuinely hard material.

• Create an API key in your OpenAI or Anthropic account and add a little credit.
• Paste it into Settings and click Test API Connection.
• Your key is stored in your OS secure keychain (macOS Keychain or Windows Credential Manager), not in some company database.

3. Start continuous and let the timer work

Hit Start Continuous and the app captures that window on a timer, every 30 seconds by default, and sends each shot to GPT or Claude for an explanation in a side panel. Speed the interval up for rapid review or slow it down for deep reading. A stability gate waits for the screen to settle before it reads, so you don't burn a capture on a half-scrolled page. Prefer one shot at a time? Use Capture Now instead, and a global stop hotkey stops capture instantly. This is built for unattended review of your own materials while you prepare, not for use during any live or proctored test.

4. Ground answers in your own notes

A general model gives general answers. Point a persona at a folder of your own notes and guides (.md, .txt, .json, or .xml; convert PDFs to text or Markdown first) and Auto-Snapper AI treats that material as authoritative on every capture. Edit the notes anytime and click Re-ingest to refresh. Built-in prompt caching means you don't pay full price to re-send your notes on each shot.

5. Read the confidence ratings

Every answer comes with a confidence rating and the reasoning behind it, so you know which explanations to trust and which to double-check.

• High: a clean answer you can rely on and move past.
• Medium: skim the reasoning the app shows before you accept it.
• Low: a clear flag that the topic is shaky, so verify it against your own materials.

Everything is local and searchable: history stays on your machine, you can replay a session, and one button clears it. Export a run as a study bundle or as Markdown, JSON, or CSV when you want a focused second pass.