How to Reply to Anything on Your Screen With One Shortcut

Every AI-assistant launch in 2026 has shipped the same headline feature: something that can see your screen, not just the text you paste into it. That's not a coincidence - manually copying context into a chat window is the single biggest source of friction in using AI day to day, and "screen-aware" is how the category is fixing it. Echoo has shipped this since early on, as Screen Context - here's how it actually works and how to set it up.

What Screen Context does (and doesn't do)

Screen Context is a per-command toggle. When it's on and you trigger that command, Echoo captures a screenshot of your display using macOS's native ScreenCaptureKit and sends it to your AI provider alongside your selected text - or on its own, if you're using voice with nothing selected. The model sees the surrounding conversation, document, or app, not just the sentence you highlighted.

A few things worth knowing before you turn it on:

• It's off by default, and it's toggled per command - not a blanket "always watching" mode.
• It requires macOS Screen Recording permission, which you'll be prompted for the first time you use it.
• The screenshot goes to whichever AI provider you've configured for that command - OpenAI, Anthropic, Google, or a local model via Ollama. It is never stored on your Mac or sent to Echoo's own servers; see the full privacy breakdown (opens in new tab) for how Echoo handles data generally.
• You need a multimodal model. Current GPT and Claude models, along with Gemini, all handle image input; a text-only model just ignores the screenshot. If you want the screenshot to never leave your device, run the command through a local multimodal model via Ollama instead of a cloud provider.

Setting it up

• Open Echoo's settings and pick a command - a built-in one like Ask, or a custom command (opens in new tab) you've made.
• Toggle Screen Context on for that command.
• Trigger the command as usual - select text and hit the shortcut, or use voice with nothing selected.
• Approve the Screen Recording permission prompt the first time; macOS only asks once.

That's the entire setup. From here it's just: press shortcut, get a reply that already knows what's on your screen.

Three things to actually do with it

Reply to a thread without re-reading it. This is the workflow behind Slack Answer (opens in new tab) - a command with Screen Context on that reads the visible messages above your cursor and drafts a reply that accounts for the whole exchange, not just the last line. Read the full story on why it stuck (opens in new tab), then build the same idea for Mail or Messages.

Summarize what's in front of you. Trigger a Summarize command with Screen Context on and no text selected, and Echoo switches into screen-only analysis mode - it reads the whole visible window (a long doc, a dashboard, a PDF you don't want to select all of) and summarizes it directly.

Translate what you're looking at, in place. Same idea with a Translate command - useful for a webpage or PDF where selecting the exact text is awkward, but the content on screen is clear.

When to leave it off

Screen Context is genuinely useful, but it's not free - it adds a screenshot to every request, which costs more tokens and a little more latency, and it means whatever's on screen goes to your AI provider. Leave it off for simple commands like grammar fixes and translation where the extra context doesn't change the output, and turn it on selectively for anything where the surrounding conversation or document actually matters - replies, summaries, and context-dependent rewrites.

Download Echoo (opens in new tab) and toggle Screen Context on your first command in under a minute.