Windows screenshot utility for AI chat

TileSnap

See something. Show something. Get an answer.

Capture the exact thing on your screen, crop out the noise, mark up what matters, redact secrets, and paste an AI-ready screenshot without opening an editor.

Avoid giant full-resolution dumps. No cloud upload required.

* Windows may warn "Unknown publisher" during install, TileSnap isn't code-signed yet. See why.

TileSnap capture toolbar
AI-ready copy
Scaled copy goes to clipboard. Original stays saved locally. Preview shown with the 1024px preset.
Up to 22.9x fewer tokens vs. a raw 4K screenshot, see the math below
30 free captures a day No signup wall, no credit card
Switch resolution after capture No redo, no re-crop, just pick another preset

Screen-to-AI workflow

Point at your screen and say: "look at this."

A broken layout, a stack trace, a weird settings panel, a game bug. TileSnap turns that moment into one clean screenshot for AI.

Clipboard-first workflow

The fastest path from screen to AI chat.

Capture like Snipping Tool

Use a hotkey, drag a region, or grab the full screen. TileSnap keeps the capture flow fast and familiar.

Paste the lighter version

Auto, 512px, 1024px, and 1536px presets resize only when needed. No artificial upscaling.

Keep the original

The full-resolution screenshot is saved in the background, so you can paste small and still recover detail later.

Copy visible text

Turn readable screen text from errors, logs, menus, and settings panels into clipboard text for AI agents and quick notes.

TileSnap preview window showing a captured dashboard bug screenshot, with a preset picker, Save as, and Copy buttons
Click the post-capture thumbnail to open this preview: switch presets, Save as, or Copy again, all without redoing the capture.

Real token math

Image models count pixels differently.

OpenAI's GPT-4o family charges high-detail images in 512px tiles. Claude scales more gradually with image size. Newer models like Gemini 3 and GPT-5.6 use automatic token-budget systems instead, so there is no single exact pixel multiplier that applies to every provider.

Example: 4K screenshot to 512px preset

Using a concrete 3840 x 2160 screenshot resized by TileSnap to 512 x 288. Estimates are based on OpenAI's and Anthropic's own published vision-token formulas.

Provider Original 4K TileSnap 512px Reduction
OpenAI high detail (GPT-4o) 1105 tokens 255 tokens 4.3x fewer
Claude current top models 4784 tokens 209 tokens 22.9x fewer

Gemini 3 and GPT-5.6 are left out of this multiplier table on purpose: both use automatic token-budget systems rather than a clean public pixel-to-token formula. Smaller screenshots still help there too, just without one exact number to quote.

* This is a best-case example, real-world savings vary with your screen size and chosen preset.

Positioning

Snipping Tool is for taking screenshots. TileSnap is for making them AI-ready.

Snipping Tool

Capture, mark up, save, and share.

TileSnap

Do all of that, then make the screenshot ready to paste into AI.

Simple pricing

Free enough to feel useful, paid when it becomes part of your day.

Free

$0

  • 30 AI-ready captures per day
  • All presets, up to 1536px
  • Crop, highlight, and arrow markup
  • Originals saved locally
Download for Windows