Genai Capture

io.github.linhnnmt由 GitHub 帳號 linhnnmt 發布v1.0.1更新於 Oct 24, 2026

Capture & annotate your screen from an AI agent: arrows, boxes, text, highlight, blur/redact.

已驗證STDIO僅桌面開發者工具媒體與設計

概覽

AI 產生的概覽

讓 AI 助理擷取螢幕、檢視影像,並用箭頭、方框、文字、螢光標示和模糊或馬賽克進行標註。

可以做什麼
桌面應用程式同時扮演本機 MCP 伺服器,提供 capture_fullscreen、capture_region、annotate、get_image 和 save_image 工具。助理可以擷取全螢幕或區域畫面、檢視產生的影像、加上箭頭、形狀、文字、螢光標示、馬賽克遮蔽與浮水印等標註,然後儲存結果。擷取時會發出提示音,處理過程都在本機進行。
適用情境
適合讓助理依一句指令產生視覺化錯誤報告、前後對照證明或附標註的截圖。適用於助理需要查看螢幕內容並加以標註的桌面工作情境。
執行需求
以本機程序在使用者電腦上執行,僅支援桌面端。需先安裝桌面應用程式(v1.2.0 以上),再透過 npm 啟動器 genai-capture-mcp 搭配 npx 加入伺服器,或直接用 --mcp 參數指向已安裝的執行檔(不需 Node)。在 macOS 上需授予螢幕錄製權限,否則擷取會是黑畫面。
安裝前請注意
此伺服器會擷取螢幕,桌面上可見的任何內容都可能傳送給所連接的 AI 用戶端;影像只會傳給該用戶端。macOS 上應用程式尚未公證,首次啟動可能需要在隱私權與安全性中允許開啟。擷取需要螢幕錄製權限。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 Genai Capture,將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Desktop only,透過 STDIO。 STDIO 服務會啟動本機處理程序,因此需要 SourceWeft 桌面主機。

其他 MCP 客戶端

參照 儲存庫 中的啟動說明。

README

GenAI Capture

[Downloads] [Latest release]

Screen capture + annotation, two ways:

  • Desktop app — capture the whole desktop, annotate (arrows, shapes, text, highlighter, mosaic redaction, watermark), then copy / save / pin. Global hotkeys, gallery, themes. Cross-platform (Java).
  • JetBrains IDE plugin (this repo's source) — the same engine inside any JetBrains IDE: GoLand, PhpStorm, PyCharm, WebStorm, IntelliJ, CLion, Rider, …

Download the desktop app

Grab the installer for your OS from Releases:

OSFile
macOSGenAI Capture-<version>.dmg
Windows (installer)GenAI Capture-<version>.exe
Windows (portable)GenAI-Capture-portable-win.zip

macOS note: the app isn't notarized yet. If macOS blocks the first launch, allow it under System Settings ▸ Privacy & Security ▸ "Open Anyway", and grant Screen Recording permission when asked. Details ship in the DMG's First Launch.txt.

The plugin is on the JetBrains Marketplace — search for GenAI Capture (or install the zip from Releases by hand).

Use with AI agents (MCP)

The desktop app (v1.2.0+) doubles as an MCP server: AI agents that speak the Model Context Protocol — Claude Code, Claude Desktop, Cursor, … — can capture your screen, look at the image, annotate it (arrows, boxes, text, highlight, blur/redact), and save the result. Visual bug reports and before/after proof, driven by one sentence to your AI.

Install the desktop app, then add the server to your client. Easiest (cross-platform, via the genai-capture-mcp launcher on npm):

bash
claude mcp add genai-capture -- npx -y genai-capture-mcp

Or point directly at the installed binary (no Node needed):

bash
claude mcp add genai-capture -- \  "/Applications/GenAI Capture.app/Contents/MacOS/GenAI Capture" --mcp

Tools: capture_fullscreen, capture_region, annotate, get_image, save_image. Captures play an audible cue; everything runs locally — images only go to the AI client you connected.


JetBrains IDE plugin

Brings the screenshot capture + annotation engine inside any JetBrains IDE (GoLand, PhpStorm, PyCharm, WebStorm, IntelliJ, CLion, Rider, …).

Features

  • Tools ▸ Take Screenshot — ⌘⇧1 (macOS) / Ctrl+Shift+1 (Win/Linux). Captures the whole desktop (IDE included) and opens the annotation overlay.
  • Tools ▸ Capture Desktop (Hide IDE) — ⌘⇧2 / Ctrl+Shift+2. Temporarily hides the IDE, captures the desktop behind it, and keeps the IDE hidden until you dismiss the overlay (ESC / save / copy / pin / close).
  • Tools ▸ Take Delayed Screenshot ▸ 3s / 5s / 10s — shows an on-screen countdown, then captures (time to switch windows or open a menu first).
  • Status-bar buttons — a camera icon (immediate capture) and a monitor icon (hide-IDE desktop capture). Toggle either via the status-bar right-click menu.
  • Annotation overlay — arrows, shapes, text, highlighter, mosaic redaction, watermark, with copy / save / pin.
  • Settings ▸ Tools ▸ GenAI Capture — a Watermark page (appearance, default text, behaviour, live preview) and a Toolbar page (choose/reorder the tools on the capture toolbar). All shortcuts are rebindable in Settings ▸ Keymap.

On macOS, grant the IDE Screen Recording permission (System Settings ▸ Privacy & Security ▸ Screen Recording) or the capture comes back black — same requirement as any screen-capture tool.

How it's built

It reuses the desktop app's code (../src/main/java + ../src/main/resources) via a Gradle Copy step that filters out the app shell and the parts that don't load inside the IDE's JBR (the osystem/ FFM code, jnativehook global-hotkey, tray, and the desktop-themed Settings panels). Thin FFM-free shims under src/main/java/io/genai/screenshot/ (osystem/AbstractOs, CaptureController, ScreenshotApp) satisfy the few OS calls the overlay makes. The capture itself uses java.awt.Robot (no native code). The plugin's own actions, settings pages, and status-bar widgets live under src/main/java/io/genai/screenshot/plugin/.

Built against the platform module only (com.intellij.modules.platform), so a single build runs in every JetBrains IDE.

Prerequisites

  • JDK 17+ (a JBR or any JDK 17+).
  • A JetBrains IDE installed (the build targets your local GoLand by default — see build.gradle.kts). To use a different IDE, edit the local("…") line, or switch to a downloaded one, e.g. intellijIdeaCommunity("2024.1").
  • Gradle: easiest is to open this jetbrains-plugin/ folder in IntelliJ IDEA (it imports the Gradle project and provides the wrapper). Or brew install gradle and run gradle wrapper once.

Run it (sandbox IDE)

bash
cd jetbrains-plugin./gradlew runIde          # launches GoLand with the plugin loaded

Then use Tools ▸ Take Screenshot (⌘⇧1) or the status-bar camera icon.

Build an installable plugin

bash
cd jetbrains-plugin./gradlew buildPlugin      # -> build/distributions/genai-capture-plugin-1.0.0.zip

Install in any IDE: Settings ▸ Plugins ▸ ⚙ ▸ Install Plugin from Disk… → pick the zip.

Notes / next steps

  • The reused overlay code is verified to compile at Java 17 with no FFM / jnativehook dependency.
  • First runIde may need minor tweaks to the IntelliJ Platform Gradle DSL or the local(...) IDE path for your exact versions — iterate from there.
  • Possible future options: capture a chosen region only, or a configurable default delay.

來源:README.md,提交 c92a5a2

工具

0
SourceWeft 尚未在沙箱中啟動過該軟體包,因此其工具未知。

版本歷史

1
  1. v1.0.1最新Sep 16, 2026

評分

已滿足適用檢查項的 50%

該評分衡量此收錄條目的完整程度,依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價,也不表示伺服器安全或目前可用:可連通性是帶日期的檢查結果,評分也不包含安裝量或使用者評價。

  • 近 30 天內可連通25 分

    請保持端點能夠回應 MCP 握手;SourceWeft 每 30 天探測一次。

  • 已宣告認證方式10 分

    請在註冊中心條目中宣告憑證請求標頭,或保持端點能夠回應探測,以便判定其認證方式。

  • 已列出工具15 分

    請讓伺服器在無憑證時回應 tools/list,或發布可在沙箱中無憑證啟動的軟體包。

  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分