PHOSPHOR APPS — AGENTIC DEV TOOLING

GLANCE

Screenshot tool for AI agents

42×
smaller file
16×
fewer tokens
0%
information lost
MacBook Pro full-screen capture (3456 × 2234)CAPTURE4.2 MB PNG~2,120 tokensGLANCE HANDLES THIScompressWebP 40% · 25%↓clipboard48 KB · 4 formats⌘Vinto ClaudePASTE48 KB WebP~96 tokens

We took a screenshot of
1,000 words
The screenshot cost fewer tokens than the words!

1,000 words as text
~1,350 tokens
Screenshot of those words via Glance
~625 tokens

A Glance screenshot of a full page of text is 2× cheaper than pasting the text itself. The image is literally cheaper than the words it contains.

Read the 1,000 words →

The AI can still read this.

Same screenshot. Same information. 97% less data.

AGENT MODE
WebP · 40% quality · 25% scale
Glance settings captured in agent mode
File size3.1 KB
Dimensions239 × 300
Tokens~96
HUMAN
PNG · 100% quality · 100% scale
Glance settings captured in human mode
File size130 KB
Dimensions956 × 1196
Tokens~1,524
42×smaller file
16×fewer tokens
0%information lost
// HOW_TOKENS_WORK

AI vision models don't see file sizes — they see pixels. The formula is simple: tokens = (width × height) / 750. A human-mode 956×1196 screenshot costs ~1,524 tokens. Glance scales it to 239×300 — just ~96 tokens. That's 16× cheaper.

Smaller dimensions = fewer pixels = fewer tokens = cheaper API calls. The AI reads both images equally well.

Different settings, same readability

Human mode, Retina Native PNG
Human · Retina Native
121 KB
~1,535 tokens
Balanced quality, Small scale
Balanced · Small
4.8 KB
~96 tokens
Pixel Perfect quality, Compact scale
Pixel Perfect · Compact
6.8 KB
~44 tokens
Pixel Perfect quality, Tiny scale
Pixel Perfect · Tiny
4.7 KB
~26 tokens

All four are perfectly legible to Claude — even the 124×156 one. The smallest uses 59× fewer tokens than the human PNG.

AGENT MODE
⌘ Shift \
  • outputWebP
  • quality40% default
  • scale25% of original
  • purposeAI consumption

“Optimized for AI consumption”

4200 KBoriginal48 KBWebP @40%↓98.9%smaller
// BEFORE → AFTER
before  screenshot.png   4.2 MB
after   capture.webp     48 KB↓ 98.9% smaller
HUMAN
⌘ Ctrl Shift \
  • outputPNG
  • quality100%
  • scaleconfigurable
  • purposehuman review

“When humans need to see it too”

// OUTPUT FORMAT
format   lossless PNG
depth    full fidelity
scale    1x — 4x options
use      review · share · archive
CLIPBOARD-FIRST

Four formats written simultaneously. Works everywhere — Slack, Figma, terminal, Claude. No export step. No save dialog.

PNGimage dataTIFFlosslessfile://URL ref/pathfilepath
MENUBAR NATIVE

Lives in your menubar. No dock icon. No window. Just there when you need it — invisible when you don't.

TOKEN CALCULATOR

See exactly how many tokens your screenshot costs before sending. Budget your context window deliberately.

CONFIGURABLE

Quality sliders, scale factors, output directories. Tune everything. Live preview with playful quality labels.

DUAL ICONS

Tray icon changes to show current primary mode at a glance. Always know what mode you're in without opening anything.

BUILDER
$7/mo
support ongoing development
  • Everything in Supporter
  • Early access to beta builds
Fund what you use. Shape what comes next.
SUBSCRIBE
PATRON
$15/mo
shape the roadmap
  • Everything in Builder
  • Priority feature requests
Direct line to what gets built.
SUBSCRIBE
// WHAT_GLANCE_SAVES_YOU
Every screenshot you send to an AI costs tokens.

Per screenshot (Claude Opus 4.6 @ $5/MTok):
  Raw:    ~2,120 tokens  $0.0106
  Glance: ~96 tokens    $0.0005

At 50 screenshots/day:
  Without Glance:  $0.53/day → $15.90/month
  With Glance:    $0.02/day → $0.72/month

  You save $15.18/month. Pays for itself in 6 days.
name>
email>
ref>
msg>