PeekPeek vs a plain screenshot
A screenshot is a picture. When you paste one into ChatGPT, Claude or Cursor, the model gets pixels and has to read the text back out of the image, guess which part matters, and work out what you meant. Whatever was on screen goes with it, keys and tokens included.
Peek sends the same capture with the work already done, on your Mac first:
- Text read on your Mac. Error messages arrive word for word, so the AI can quote them exactly.
- Secrets removed first. API keys, tokens, passwords, card numbers and emails are masked in the text and blurred in the image before anything is sent.
- Written for where you paste. Prose for chat apps, Markdown for AI editors, compact text for terminal agents, with the project and git branch they're working in.
- Your intent, not just pixels. With Point & Talk, Peek tells the AI what each mark covers and what you said while drawing it.
- No file juggling. Coding agents can fetch captures themselves through Peek's MCP server and the
peekcommand.
When a plain screenshot is enough
If you only need to show someone what a screen looks like, a plain screenshot is the simpler tool. Peek is for the moment you want an AI to understand the screen and act on it.