Docs

Screen & Context

Agentastic is most useful when it can see what you see. You decide what context to share — nothing is captured until you ask.

Screenshots and screen capture

Capture part of your screen and hand it to the agent:

  • Press the screenshot shortcut (default ⌘4, also ⌘⇧S), then drag to select a region. Esc cancels.
  • The image is attached to your conversation and Vision mode turns on automatically so the model can analyze it.
  • The captured image is also copied to your clipboard.

Screen capture requires Screen Recording permission (macOS 14+).

Choosing a display

On a multi-monitor setup, use the /screen: tag to pick which display to capture:

  • /screen:active or /screen:0 — the current display
  • /screen:1, /screen:2, … — a specific monitor by number

Browser tabs

Bring the contents of open browser tabs into the conversation:

  • Type /tab: at the end of your prompt to open the tab selector.
  • Select one or more tabs (⌘-click, Shift-click for a range). Supported browsers include Chrome, Arc, and Safari.
  • Agentastic extracts each tab's content and adds it as context, so you can ask things like "compare these two product pages."

App context

When you open the launcher, Agentastic knows which app was in front and can read your selected text. That lets you select a paragraph in any app and ask the agent to rewrite, translate, or explain it without copying anything.

Working directory

For coding and file tasks, set a working directory so the agent reads and writes the right files. Agentastic also looks for instruction files in that folder — AGENTS.md, AGENTASTIC.md, CLAUDE.md, or agent.md — and follows the guidance they contain, the same way it would in a code project.

Putting it together

/agent Look at /screen: and the error in my active tab /tab:,
then explain what's failing and propose a fix.

This captures your screen, pulls in the selected browser tab, and runs an agent that reasons over both.

Related