Shotik
Open-source screenshot tool for Windows with a built-in MCP server that allows AI assistants like Claude to capture and view screen content through tools like take_screenshot, ask_user_to_select_region, and more.
README
<p align="center"> <img src="assets/icon.png" width="96" alt="Shotik"> </p>
<h1 align="center">Shotik</h1>
<p align="center"> <b>Screenshots for humans & AI</b> · a beautiful open-source screenshot tool for Windows<br> with a built-in MCP server — Claude can see your screen </p>
<p align="center"> <a href="README.ru.md">Русская версия →</a> </p>
<p align="center"> <a href="https://github.com/gorka2354/shotik/releases/latest"><img src="https://img.shields.io/github/v/release/gorka2354/shotik?label=release&color=2ea44f" alt="Release"></a> <a href="https://github.com/gorka2354/shotik/releases"><img src="https://img.shields.io/github/downloads/gorka2354/shotik/total?color=blue" alt="Downloads"></a> <img src="https://img.shields.io/badge/platform-Windows%2010%2F11-0078d4" alt="Platform"> <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-lightgrey" alt="MIT"></a> <img src="https://img.shields.io/badge/MCP-built--in-7c5cff" alt="MCP"> </p>
<p align="center"> <img src="docs/demo.gif" width="88%" alt="Demo: window hover-snap, annotations, copy"> </p> <p align="center"><i>hover highlights a window → click → arrow, box, pixelate, markers, text → Enter → in your clipboard</i></p>
<p align="center"> <img src="docs/app-dark.png" width="46%" alt="Dark theme"> <img src="docs/app-light.png" width="46%" alt="Light theme"> </p> <p align="center"><i>the UI follows your system theme and Windows accent color · English & Russian interface</i></p>
Why another screenshot tool?
Shotik is ShareX-style region capture with annotations, built around two ideas:
- AI-first. Screenshots today often go to Claude/ChatGPT in a console, not to people. Shotik makes that zero-friction.
- Speed without dialogs. Capture → already in the clipboard, already on disk, ready to paste.
Features
✦ Smart Clipboard
After every capture the clipboard holds both formats at once: the PNG image and the file path as text.
Ctrl+V into Claude Code pastes the image; Ctrl+V into a plain terminal pastes the path. No switching.
✦ Built-in MCP server
Connect once:
claude mcp add shotik --transport http http://127.0.0.1:7464/mcp
and Claude gets these tools:
| Tool | What it does |
|---|---|
take_screenshot |
look at the screen right now |
ask_user_to_select_region |
open the overlay so you point at the area |
get_last_screenshot |
grab your latest shot |
take_screenshot_region |
capture a specific rectangle |
list_screenshots, list_displays |
history and monitors |
Say “look at my screen” — and Claude sees it. Say “look here” — and just draw a box with your mouse.
✦ Repeat last area (Shift+PrtSc)
Same region, fresh pixels, instant, no overlay. Perfect for iteration loops:
fix the CSS → Shift+PrtSc → Ctrl+V to Claude → “see it now?”. Survives restarts.
✦ Flameshot-style overlay with annotations
Freeze-frame, magnifier with pixel-perfect HEX picker, boxes, arrows, pen, highlighter,
pixelate your secrets, text, numbered markers — all inside the overlay, before saving.
Hovering highlights the window under the cursor; click captures it (hold Alt to disable snapping).
✦ PowerToys Command Palette / Win+S
The installer adds Start Menu commands that show up in PowerToys Command Palette and Windows search:
Shotik Area Screenshot, Shotik Full Screen, Shotik Repeat Last Area.
CLI works too: Shotik.exe --capture region|full|repeat.
✦ And the rest
- Pin a shot on top of every window exactly where you cut it (wheel = zoom, right-click = menu)
- OCR the selection via built-in Windows OCR (offline, your installed languages)
- Color picker —
Cin the overlay copies the pixel HEX - History gallery, toasts with previews, multi-monitor, HiDPI
- Native look: light/dark theme and accent color come from the OS and switch live
- English & Russian UI (follows the system language, switchable in Settings)
Hotkeys
| Action | Windows | macOS |
|---|---|---|
| Capture area | PrtSc |
⌘⇧2 |
| Full screen (monitor under cursor) | Ctrl+PrtSc |
⌘⇧1 |
| Repeat last area | Shift+PrtSc |
⌘⇧7 |
In the overlay: Enter/double-click — copy · Ctrl+S — save as · P — pin · T — OCR ·
A — copy for Claude · F — whole screen · 1..0 — tools · [ ] — stroke width ·
Ctrl+Z — undo · arrows — nudge selection · Esc — cancel.
Install
Prebuilt: Releases →
Shotik-Setup-x.x.x.exe (installer with shortcuts) or Shotik-x.x.x-portable.exe (single file, no install).
From source:
git clone https://github.com/gorka2354/shotik && cd shotik
npm install
npm start
Windows 10/11 required (Node.js ≥ 20 for building from source). The app lives in the tray; closing the window minimizes it, quit via the tray menu.
macOS (experimental): .dmg builds (arm64 & x64) are produced in Releases. Default hotkeys
are ⌘⇧2 / ⌘⇧1 / ⌘⇧7; OCR uses Apple Vision; window hover-snap is Windows-only for now.
Builds are unsigned: right-click → Open on first launch, and grant Screen Recording permission.
The maintainers have no Mac — feedback and PRs are very welcome.
Microsoft Store: packaging is ready (npm run dist:store builds the MSIX), pending a
Partner Center listing — see docs/STORE.md.
Monitors: multi-monitor and mixed resolutions are fully supported (capture and overlay are computed per display, HiDPI-aware). Known limitation: with mixed DPI scale factors the window hover-snap highlight can be a couple of pixels off.
For contributors: ghost mode
E2E tests never touch your desktop: windows are placed off-screen, the desktop is replaced by a
fixture image, global hotkeys are not registered, and everything is driven over HTTP /test/*
endpoints:
$env:SHOTIK_TEST='1'; $env:SHOTIK_GHOST='1'
$env:SHOTIK_FAKE_SCREEN='test\fake-screen.png'
npm start -- --hidden
# POST http://127.0.0.1:7464/test/trigger {"mode":"region"}
# POST http://127.0.0.1:7464/test/input {"events":[...]} — synthetic input
# POST http://127.0.0.1:7464/test/capture-page — window UI snapshot
The README demo GIF is robot-filmed too: node test/film-cinematic.js drives the cursor with
eased steps (~200 frames + a manifest), and node test/render-camera.js applies a virtual
follow-camera (lazy cursor tracking + push-ins) and encodes the GIF.
Simple storyboard variant: film-demo.js + make-gif.js.
Architecture
src/
main/ main process: capture (desktopCapturer), history, OCR (WinRT / Apple Vision),
MCP server (zero-deps streamable HTTP), windows, settings, i18n
overlay/ selection overlay: canvas rendering, annotations, toolbar
app/ main window: gallery, Claude page, settings
pin/ toast/ pinned shots and notifications
The only runtime dependency is Electron. No bundlers, no frameworks, no native modules.
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.