Capture your screen,
send the screenshot straight to your AI agent
Screenshot to Agent puts a screenshot button right in the AI agent composer and gives you a configurable keyboard shortcut to grab a screen region from anywhere. Press the shortcut or click the button, drag a box around what you want, and the screen capture attaches itself to your next message. The screenshot lands next to your prompt and goes straight to your AI coding agent.
No external screenshot tool, no saving a PNG to your desktop, no hunting for the file to drag it into the window. You skip the whole capture-save-locate-drop dance and save a handful of seconds every single time, which is the whole point: tighter, higher-fidelity communication between you and your agent.
Screenshot to Agent in action: hit the keyboard shortcut, drag a box around the screen region, and the capture is auto-attached to your AI agent's message.
Here is the workflow Screenshot to Agent kills. You are vibe coding with an AI agent and something on screen is off, a misaligned card, a wrong color, a console error, a layout that breaks at a certain width. To show the agent, you reach for a separate screenshot tool, drag a region, save a PNG somewhere, find that file, then drag it into the composer. Five steps, a context switch, and the conversation with your agent is on pause the whole time.
Screenshot to Agent collapses all of that into one move. You press a keyboard shortcut, your operating system's own region picker appears, you drag a box, and the screenshot is captured and attached to the agent composer automatically. No external app to open, no file to save, no file to locate, no drag and drop. The screen capture is sitting next to your prompt, ready to send.
It is two entry points for the same thing. A screenshot button in the composer when your hands are already on the mouse, and a customizable global shortcut when you want to grab the screen from anywhere without clicking around first. Either way the result is identical: a screenshot, attached to your message, on its way to your AI coding agent in seconds.
Two ways to send a screenshot
A button when your hand is on the mouse, a shortcut when it is on the keyboard. Same result.
The screenshot button in the composer
Every agent composer has a screenshot button. Click it and your operating system's region picker opens so you can drag a box around any part of your screen. The capture is saved and attached to that agent's composer as a thumbnail, ready to send. It is the obvious path when you are already in the composer and reaching for the mouse.
The configurable keyboard shortcut
A global keyboard shortcut, Cmd+Shift+2 by default on macOS (Ctrl+Shift+2 on Windows and Linux), triggers the same region capture from anywhere, even when AgentsRoom is not the focused window. The screenshot is routed to your active agent's composer automatically. The shortcut is fully customizable: rebind it to whatever combo you like, or turn it off, in Settings.
Capture a screen region with the shortcut or the composer button, and the screenshot is auto-attached to your AI agent's message as a thumbnail, ready to send with your prompt.
Why Screenshot to Agent belongs in your agent workflow
Speed. Capturing a screen region and getting it in front of your agent used to be five steps across two apps. Now it is one shortcut and a drag. You save a few seconds on every screenshot, and when you screenshot the agent dozens of times a day, those seconds add up to real flow.
Precision. A screenshot shows the agent exactly what you see, pixel for pixel, instead of a paragraph trying to describe it. The wrong padding, the broken alignment, the error in the corner, the agent gets the real picture, not your best attempt at narrating it.
No external tool. You do not open a separate screenshot app, you do not litter your desktop with PNG files, you do not dig through Finder to find the one you just took. The capture and the attach happen in a single action, inside AgentsRoom.
Flow. The shortcut works from anywhere and the screenshot comes back into the composer next to your prompt. You never leave the terminal, you never break the conversation with the agent, and your screen capture and your written instructions travel together in the same message.
How Screenshot to Agent works, end to end
From a keyboard shortcut to a screenshot attached to your agent's message.
Trigger the capture
Press the keyboard shortcut (Cmd+Shift+2 by default) from anywhere, or click the screenshot button inside the agent composer. Both start the exact same region capture.
Drag a box around the region
Your operating system's own region picker takes over: a crosshair on macOS via screencapture, the Snip overlay on Windows, grim, spectacle or gnome-screenshot on Linux. Drag a box around the part of the screen you want. It is the capture UI you already know.
The screenshot attaches itself
The captured region is saved as a PNG and dropped straight into the focused agent's composer as a thumbnail, exactly like a file you dragged in. No saving to your desktop, no locating the file, no drag and drop. It is already there.
Write your prompt around it
Type your message as usual. Reference the capture in your words if you want, or just let the attached screenshot speak for itself. The screenshot shows up as a numbered file in the composer, so you can point the agent at it explicitly.
Send it to the agent
Hit send. The PNG goes out with your message through the same pipeline as any other attachment. Your screen capture and your written instructions reach the agent together, in a single turn.
The agent reads your screen
Your AI coding agent receives the screenshot as an attached image. As long as its model accepts image input, it sees exactly what was on your screen and acts on it, instead of parsing a paragraph that tries to describe a visual problem.
What Screenshot to Agent really is
Screenshot to Agent is a region capture flow wired directly into the AgentsRoom desktop app. It is not a screenshot viewer bolted on the side. It hooks into your operating system's native region picker, the same screencapture crosshair on macOS, the same Snip overlay on Windows, the same grim, spectacle, gnome-screenshot, maim, scrot or flameshot tools on Linux, so the capture experience is the one you already use, with nothing new to learn.
The two triggers, the composer button and the global keyboard shortcut, both run the same capture and both end the same way: the screenshot is attached to the focused agent's composer. The shortcut is global, so it fires even when AgentsRoom is in the background, and it is fully configurable. You rebind it or disable it from the composer settings, where a shortcut recorder lets you press the exact combination you want.
Once captured, the PNG is saved locally and added to the composer's file tray as a thumbnail, where it picks up a reference number like any dropped file. From there it follows the normal path: it goes out with your prompt and is sent to the agent as an image. Screenshot to Agent does not invent a new sending mechanism, it plugs into the one that already moves files to your agent.
Everything stays on your machine. Captures are saved to a local folder in your home directory, not inside your project repo and not on any third-party service, and they are cleaned up automatically so they never pile up. Screenshot to Agent is local, private and provider-agnostic: it attaches a plain PNG, so any agent whose model accepts an image can read it, which means Claude Code and Codex setups get the same capture-to-agent flow. It is a desktop feature, available on macOS, Windows and Linux.
What people screenshot to their agent
Three moments where a screenshot beats a paragraph of description.
Show a visual or UI bug
Capture the broken interface, the misaligned element, the wrong spacing, the console error, and send it. The agent sees the actual bug instead of decoding your description of it. The fastest way to report a UI regression to an AI coding agent.
Point at what is on screen
A chart, a terminal output, a diff, a third-party dashboard, anything visible on your screen. Capture the region and hand it to the agent as context instead of retyping what it says. The screenshot carries the detail your words would drop.
Pass a visual reference
A design you are matching, a layout you like, a component from another app. Grab a screenshot of the reference and give it to the agent as a target, then let it build toward what you captured.
Describe it, or just screenshot it
Same screen, same agent. One path is a five-step chore, the other is a shortcut.
Without Screenshot to Agent
- : You open a separate screenshot tool, drag a region, and save a PNG somewhere.
- : You hunt for that file, then drag and drop it into the composer.
- : Or you give up and try to describe the visual problem in words.
- : The context switch breaks your flow and pauses the conversation with the agent.
- : Every screenshot costs you the same handful of seconds, all day long.
With Screenshot to Agent
- : You press the shortcut or click the button and drag a box. One move.
- : The screenshot attaches itself to the composer automatically.
- : No external app, no saved file, no locating it, no drag and drop.
- : You never leave AgentsRoom and never break the thread with the agent.
- : The capture and your prompt reach the agent together, in one turn.
Screenshot to Agent is the shortest path between what is on your screen and an AI agent that can see it.
FAQ
What is Screenshot to Agent in AgentsRoom?
Screenshot to Agent is a built-in way to capture a region of your screen and send it straight to your AI coding agent. You trigger it from a screenshot button in the agent composer or from a configurable global keyboard shortcut, drag a box with your operating system's region picker, and the screen capture is auto-attached to your next message. No external screenshot tool and no moving files.
How do I send a screenshot to an AI agent?
Press the keyboard shortcut (Cmd+Shift+2 by default, Ctrl+Shift+2 on Windows and Linux) or click the screenshot button in the composer, then drag a box around the part of your screen you want. The screenshot is captured and attached to the agent composer as a thumbnail. Write your prompt and send: the capture goes to the agent together with your text.
What is the default keyboard shortcut, and can I change it?
The default is Cmd+Shift+2 on macOS and Ctrl+Shift+2 on Windows and Linux. It is a global shortcut, so it fires even when AgentsRoom is not the focused window. You can rebind it to any combination you like, or turn it off entirely, in the composer settings, where a shortcut recorder lets you press the exact keys you want.
Does it work when AgentsRoom is in the background?
Yes. The keyboard shortcut is registered globally, so you can capture a screen region from any app and the screenshot is routed to your active AgentsRoom agent's composer. You do not have to bring the window to the front first.
Do I need an external screenshot tool?
No, that is the whole point. Screenshot to Agent uses your operating system's own region picker (screencapture on macOS, the Snip overlay on Windows, grim, spectacle or gnome-screenshot on Linux) and attaches the result for you. You do not open a separate screenshot app, you do not save a PNG to your desktop, and you do not drag a file into the window.
Where are my screenshots saved? Are they private?
Captures are saved to a local folder in your home directory on your own machine. They are not committed to your project repo and not uploaded to any third-party service. They are transient by design and cleaned up automatically, so they do not pile up over time.
Which agents and providers work with Screenshot to Agent?
It attaches a standard PNG image, so it is provider-agnostic. Any agent whose model accepts image input works: Claude Code, Codex and the other CLIs AgentsRoom supports. There is no provider-specific configuration to do.
Can I capture just part of the screen?
Yes. The region picker lets you drag a box around exactly the part of the screen you want, so you send just the relevant area instead of a full-screen capture. You crop at capture time, with the native selection UI you already know.
Can I annotate the screenshot before sending it?
Screenshot to Agent attaches the raw capture. If you want to circle a bug, draw an arrow or highlight a region on top of it, open the capture in Sketch, the drawing canvas in the composer, annotate it, and the marked-up version goes to the agent. Capture with Screenshot to Agent, mark it up with Sketch.
Why screenshot the problem instead of describing it?
Because a description is a guess and a screenshot is not. When you write 'the button in the top right is misaligned', the agent has to rebuild your screen in its head. When you send the actual capture, there is nothing to rebuild. A screenshot carries the real visual detail, so the agent acts on what you see instead of on your best attempt to narrate it.
Goes well with
Sketch
A drawing canvas in the composer. Capture the screen with Screenshot to Agent, then open it in Sketch to circle the bug, draw an arrow or highlight a region before sending.
Voice Dictation
Dictate your prompt instead of typing it. Pair a spoken instruction with the screenshot you just captured and the agent gets both at once.
Prompt Library
Save and reuse your best prompts. Fire a saved prompt next to a fresh screen capture and skip the typing entirely.
Scratchpad
A notepad for prompts and notes in AgentsRoom. Draft the words, attach the screenshot, send the whole context to your agent composer.
Stop saving PNGs. Just screenshot it.
Download AgentsRoom and use Screenshot to Agent: a screenshot button and a configurable keyboard shortcut that capture a screen region and send it straight to your AI coding agent, with no external screenshot tool and no moving files.
Companion app: monitor your agents on the go
Bring your own: Claude, Codex, Gemini CLI, or other AI provider.
Push bugs and requests straight to your public backlog.
A glimpse of AgentsRoom in action.