A Claude and coding-agent skill that turns a short JSON spec into animated, explorable diagrams exportable as SVG, PNG, GIF or MP4.
rixinhahaha/snip
Electron menu bar app that lets AI coding agents render diagrams and previews, and lets you annotate and approve them on screen.
Open-source alternative to:CleanShot X
- Stars
- 347
- Forks
- 32
- Open issues
- 4
- Last commit
- 6 months ago
What is it?
Snip is a visual review layer for AI coding agents. Once installed, a command such as snip setup configures Claude Code (and Cursor, Windsurf, Cline) so that architecture, data models and flows are rendered as Mermaid diagrams or HTML previews instead of text. The user reviews the output on screen, either approving it or annotating it with circles, arrows and notes, and the agent receives structured JSON with the decision and annotations.
Snip also works as a standalone screenshot and annotation tool. Capture shortcuts support region select and window snap, a quick mode copies to the clipboard, and the editor offers rectangle, arrow, text, blur and AI segmentation tools. Saved screenshots are named, categorized and tagged by a local vision model run through Ollama, and can be found with semantic search.
The project is a CLI plus an MCP server for agents without shell access. It runs on macOS 14+ and Linux under Wayland, and its core AI features run locally without cloud APIs.
Key features
- 01Renders Mermaid diagrams and HTML previews for AI coding agents
- 02Annotate screenshots with shapes, arrows and text, then send feedback back to the agent
- 03Configures Claude Code, Cursor, Windsurf and Cline via snip setup
- 04Screenshot capture with region select, window snap and clipboard quick copy
- 05Local AI categorization and semantic search of saved screenshots through Ollama
- 06Includes an MCP server for agents without shell access
Use cases
- Reviewing architecture diagrams or data models that Claude Code generates before accepting them
- Marking up UI previews with feedback that the agent can act on in the next iteration
- Searching a large personal screenshot library by describing what was on screen
Quick start
- 1
Install on macOS with Homebrew
brew install --cask rixinhahaha/snip/snip - 2
Configure Claude Code to use visual output
snip setup
Things to know
- Requires macOS 14+ or Linux (Wayland), so Windows is not supported.
- macOS builds are Apple Silicon DMG only in the README; Linux is available as AppImage or deb.
- Optional AI organization needs a local Ollama vision model installed.
Topics
FAQ
- What is snip?
- Electron menu bar app that lets AI coding agents render diagrams and previews, and lets you annotate and approve them on screen. Snip is a visual review layer for AI coding agents. Once installed, a command such as snip setup configures Claude Code (and Cursor, Windsurf, Cline) so that architecture, data models and flows are rendered as Mermaid diagrams or HTML previews instead of text. The user reviews the output on screen, either approving it or annotating it with circles, arrows and notes, and the agent receives structured JSON with the decision and annotations.
- Is snip free to use? What license does it have?
- snip is open source under the MIT license. Check the license terms before using it in commercial products.
- Is snip still maintained?
- Updated within 6 months. The last commit to snip was on May 7, 2026.
- What can I use snip for?
- Reviewing architecture diagrams or data models that Claude Code generates before accepting them Marking up UI previews with feedback that the agent can act on in the next iteration Searching a large personal screenshot library by describing what was on screen
- How do I install snip?
- Install on macOS with Homebrew: brew install --cask rixinhahaha/snip/snip Configure Claude Code to use visual output: snip setup
- Where can I get snip?
- The source code and installation instructions are on GitHub: https://github.com/rixinhahaha/snip
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Desktop app for browsing, installing and managing AI agent skills across 20+ coding agents from one interface.
A local browser-based review surface that lets you annotate agent plans, markdown, HTML and code diffs, then send feedback back to coding agents.
A macOS app for browsing, editing, and organizing AI coding agent skills and agents across Claude Code, Cursor, Codex, Windsurf, and Amp.
A skills framework and software development methodology that guides coding agents through brainstorming, planning, TDD, and reviewed subagent-driven implementation.
Persistent memory for coding agents that captures session activity, compresses it with AI, and injects relevant context into future sessions.