Visual AI Tool for macOS & Linux

Stop describing
what you see.

A screenshot with an arrow beats a paragraph of explanation. Snip is the visual channel between you and AI.

Download for macOS (Apple Silicon) View Source
$ brew install --cask rixinhahaha/snip/snip

Requires macOS 14+ with Apple M-series chip (M1, M2, M3, M4). Free and open source.

Snip on Product Hunt
Setup Guide
AI Workflow

Show, don't tell.
Review, don't imagine.

Circle a bug and paste to your agent — one screenshot replaces a paragraph. Or let your agent render a diagram and get your spatial feedback. The snip CLI and MCP server let any AI agent use Snip as their visual I/O.

snip open <file> Open for review — approve, annotate, or request changes
snip render --format mermaid Render a diagram from code, open for review
snip search "query" Semantic search across your library
snip transcribe <file> Extract text via native OCR (macOS only)
snip list Browse all saved snips with metadata
snip organize <file> AI categorization with local LLM
Quick Setup

Works with Claude Code, Cursor, Windsurf, Cline, and any AI agent that can run shell commands.

  1. Open Snip Settings → Install CLI
  2. Click Configure next to your detected AI tool
  3. Your AI can now call snip open, snip search, etc.

For clients that can't run CLI commands (e.g., Claude Desktop), Snip also supports MCP. Enable it in Settings and copy the config.

How It Works

Two directions. One tool.

You → Agent

Show what you see

1

Capture

Cmd+Shift+2 — select a region or click a window.

2

Annotate

Circle the bug, arrow the fix, type what's wrong.

3

Copy

Press Esc — annotated image on your clipboard.

4

Paste to agent

Ctrl+V into Claude Code, Cursor, or any AI chat. The agent sees exactly what you see.

Agent → You → Agent

Review what the agent built

1

Agent renders

Agent calls snip render or snip open with a diagram or image.

2

Review Mode opens

Snip pops up with the visual and the agent's context message.

3

Annotate & respond

Draw on the image, type feedback, or just approve.

4

Agent acts

Your annotations and text return to the agent as structured feedback.

More Features

AI organization, segmentation, upscale, animate, transcribe & more

Optional AI-powered add-ons: auto-organize with a local vision LLM, semantic search, object segmentation, 2x upscaling, text extraction, GIF animation. All on-device, all optional.
Extensions

Built to extend.

Every tool in Snip is an extension — rectangle, arrow, text, segment, upscale, transcribe. Add new capabilities by dropping a folder, or let AI agents install sandboxed extensions via MCP.

Source Extensions Add a folder to src/extensions/ with a manifest. Zero other files touched.
User Extensions Install from Settings or via MCP. Sandboxed child process, blocked modules, user approval required.
Permission System Extensions declare what they need. screenshots:read, temp:write — shown before install.
Snip Browser NEW

Your favorite annotation tools, now in the browser.

Capture, annotate, and share screenshots from Chrome. Comment tool sends visual directions as coordinates to your AI.

Capture & Annotate Select any region, annotate with arrows, text, rectangles, tags, blur, and crop. Copy or download as PNG.
AI Comments Place numbered arrow annotations with text. Comments are sent as JSON coordinates — not baked into the image — so your AI knows exactly what you're pointing at.
Works with your tools Integrates with Claude Code, Codex CLI, and Cursor via snip-wait. AI requests a capture, you annotate, it gets the image + comments back.
Get Snip Browser

Ready to snip smarter?

Free, open source, and built for power users.

Download for macOS (Apple Silicon) View on GitHub

Requires macOS 14+ with Apple M-series chip.