+ How It Works — Cai Docs

How It Works

The Core Flow

  1. Select text anywhere on your Mac
  2. Press ⌥C (Option+C)
  3. Cai detects the content type and shows relevant actions
  4. Pick an action with ↑↓ arrows or ⌘1–9
  5. The result is auto-copied to your clipboard — just ⌘V to paste. Or send it to an output destination like Slack, Bear, or a webhook.

Content Types & Actions

Cai automatically detects what you copied and offers the right actions:

Content TypeDetectionActions
URLhttps://..., www.Open in Browser
JSONValid JSON object or arrayPretty Print
MeetingDate/time referencesCreate Calendar Event, Open in Maps
AddressStreet patterns, “at [Place Name]“Open in Maps
Word1–2 wordsDefine, Explain, Translate, Search
Short TextLess than 100 charactersExplain, Reply, Fix Grammar, Translate, Search
Long Text100+ charactersSummarize, Reply, Fix Grammar, Translate, Search
ImageClipboard image or screenshotExtract Text (OCR), then all text actions

All text types also get Ask AI (⌘1) for free-form AI prompts.

Context Awareness

Cai is context-aware — it detects what app you’re in (email, Slack, code editor, etc.) and tailors its actions accordingly. For example, Reply uses this context to generate an appropriate response based on whether you’re replying to an email, a chat message, or a code review comment.

Examples

  • Select "serendipity" → Define, Explain, Translate, Search
  • Select an email body → Reply, Summarize, Translate
  • Select "Let's meet Tuesday at 3pm at Starbucks" → Create Calendar Event, Open in Maps
  • Select "123 Main St, NYC 10001" → Open in Maps
  • Select https://github.com/... → Open in Browser
  • Select {"name": "John"} → Pretty Print JSON

System Actions vs AI Actions

System actions work without any LLM server:

  • Open URL in browser
  • Open address in Maps (Apple or Google)
  • Create calendar event (via ICS file — works with any calendar app)
  • Web search
  • Pretty print JSON
  • Extract text from images (OCR via Apple Vision)

AI actions use the built-in model by default, or a connected external provider:

  • Summarize / Explain
  • Reply (context-aware responses)
  • Fix Grammar
  • Translate
  • Define
  • Ask AI (your own prompt)

Follow-up Questions

After any AI action returns a result, press Tab to ask a follow-up question. The full conversation history is sent to the LLM, so it has context from the previous result. You can chain multiple follow-ups in a single session.

Press ⌘N to start a new chat — ask anything without clipboard content.

Clipboard History

Press ⌘0 in the action window to open clipboard history. Features include:

  • Search — type to filter your history
  • Pinning — pin important items so they stay at the top
  • Configurable size — adjust the history limit in settings

Settings

Left-click the Cai menu bar icon to access Preferences:

SettingDescriptionDefault
Translation LanguageTarget language for translationsEnglish
Search URLBase URL for web searchesBrave Search
Maps ProviderApple Maps or Google MapsApple Maps
Model ProviderBuilt-in, LM Studio, Ollama, Cloud, or CustomBuilt-in
About YouPersonal context for AI responses (e.g. profession, tone)Empty
Launch at LoginStart Cai automaticallyOn