AI Context Menu — Mac App

A system-wide macOS right-click menu that instantly routes selected text, images, files, and screenshots to your preferred AI services.

---

Overview

AI Context Menu lives in the menu bar and injects itself into every right-click anywhere on macOS. Highlight text in a PDF, right-click a photo in Finder, or right-click an empty desktop — and a curated submenu of AI destinations appears, ready to send your content in one click.

---

Features

Context-aware actions

Context	What you can do
Selected text	Ask, explain, translate, summarize, rewrite
Image / media file	Describe, analyze, extract text (OCR)
Any file	Summarize, review, convert
Empty area / desktop	Capture active window → send to AI

System-wide integration

Works everywhere — browsers, native apps, Finder, Terminal, PDF viewers, code editors — via macOS Accessibility and Services APIs.

Configurable menu

Choose exactly which AI destinations appear in the right-click menu. Reorder them by drag-and-drop in Settings. Hide services you never use.

---

Supported AI Services

Cloud / Web

Service	Send method	Notes
Claude	Native app or web (claude.ai)	Supports text + images
ChatGPT	Native app or web (chatgpt.com)	Supports text + images
Gemini	Web (gemini.google.com)	Supports text + images
Perplexity	Web (perplexity.ai)	Best for search-style queries
Grok	Web (grok.com)	xAI
Mistral Le Chat	Web (chat.mistral.ai)
Copilot	Web (copilot.microsoft.com)

Local / Self-hosted

Service	Default endpoint	Notes
Ollama	http://localhost:11434	Auto-detects running models
LM Studio	http://localhost:1234	OpenAI-compatible
Jan	http://localhost:1337	OpenAI-compatible
Open WebUI	http://localhost:3000	OpenAI-compatible
AnythingLLM	http://localhost:3001	OpenAI-compatible
llamafile	http://localhost:8080	OpenAI-compatible
LocalAI	http://localhost:8080	OpenAI-compatible
text-generation-webui	http://localhost:5000	OpenAI-compatible
Custom endpoint	User-defined	Any OpenAI-compatible API

---

Custom Endpoint

For self-hosted or enterprise models, configure a custom endpoint with full control:

Base URL:   http://192.168.1.10:8080/v1
API Key:    (optional)
Model:      llama3.2

A built-in endpoint library lets you pick from common presets:

Preset	URL template
Ollama	http://localhost:11434/api
LM Studio	http://localhost:1234/v1
Jan	http://localhost:1337/v1
Open WebUI	http://localhost:3000/openai/v1
AnythingLLM	http://localhost:3001/api/openai
llamafile	http://localhost:8080/v1
LocalAI	http://localhost:8080/v1
Groq Cloud	https://api.groq.com/openai/v1
Together AI	https://api.together.xyz/v1
Fireworks AI	https://api.fireworks.ai/inference/v1
OpenRouter	https://openrouter.ai/api/v1

---

Design

Design language: Apple Human Interface Guidelines — feels like a first-party Apple utility.

Principles

• Invisible until needed. Lives in the menu bar as a small icon. No windows open unless you open Settings.
• Zero friction. Right-click → AI destination → done. No copy-paste, no switching apps.
• Clarity over options. The context menu shows only what makes sense for the selected content.

Visual language

• SF Pro for all text
• SF Symbols for all icons (brain.head.profile, doc.on.doc, camera.viewfinder, etc.)
• macOS vibrancy / .sidebar materials for panels
• Accent color inherits the user's System Accent Color
• Supports Light, Dark, and Auto modes
• Menu items match native macOS menu appearance exactly — no custom drawing in the right-click menu itself

---

Settings App

Accessible via menu bar icon → Settings… (⌘,).

Tabs

Services

• List of all available AI services
• Toggle each on/off (checkmark)
• Drag to reorder within the right-click menu
• Click a service row to expand inline configuration (API key, default model, preferred interface: web vs. native app)

Custom Endpoints

• Add / edit / delete custom endpoints
• Endpoint library browser with one-click presets
• Test connection button (shows latency + model list if OpenAI-compatible)
• Model picker populated from /v1/models response

Actions

• Define what appears in the submenu for each content type
• Example: for selected text → Ask, Explain, Summarize, Translate, Rewrite, Fix Grammar
• Custom prompts: add your own actions with a configurable system prompt template ({{content}} placeholder)

Appearance

• Menu bar icon style: color / monochrome / hidden
• Show service icons in right-click menu (on/off)
• Max items shown before "More…" overflow

Privacy

• Content is never stored or logged by the app
• Transmission goes directly to the selected service — no proxy
• Screenshot capture requires explicit Screen Recording permission

---

Permissions Required

Permission	Purpose
Accessibility	Detect and read selected text system-wide
Screen Recording	Capture active window screenshot
Automation (per-app)	Open Claude/ChatGPT native apps with content

All permissions are requested with a clear explanation on first use. The app works with only the permissions the user grants — features requiring missing permissions are gracefully hidden.

---

Right-Click Menu Structure

── Ask AI                         ▶  (appears when text is selected)
   ├── Claude
   ├── ChatGPT
   ├── Gemini
   ├── Perplexity
   └── My Local Llama              (custom endpoint)

── Send to AI                     ▶  (appears when file/image is selected)
   ├── Claude
   ├── ChatGPT
   └── Gemini

── Screenshot → AI                ▶  (appears when right-clicking empty area)
   ├── Capture active window
   └── Capture selection…

Items shown depend on:

1. Which services are enabled in Settings
2. Whether the service supports the content type (e.g., image support)
3. User-defined order from Settings

---

Technical Architecture

Layer	Technology
Language	Swift 6
UI	SwiftUI + AppKit where needed
Right-click injection	NSServicesProvider + CGEventTap fallback
Accessibility	AXUIElement API
Screenshot	ScreenCaptureKit
Local AI comms	URLSession with OpenAI-compatible REST
Web AI delivery	Deep links + URL schemes / NSWorkspace
Settings storage	SwiftData (local, no iCloud sync by default)

Minimum macOS: Sonoma (14.0)

---

Installation

1. Download AIContextMenu.dmg
2. Drag to Applications
3. Launch — grant Accessibility permission when prompted
4. Right-click anything to start

No subscription required for local AI usage. Cloud services use your own accounts.

---

Roadmap

• iCloud sync for settings across Macs
• Keyboard shortcut to send clipboard → AI without right-clicking
• Siri Shortcut / Shortcuts.app integration
• Per-app service filtering (e.g., only show coding assistants in Xcode)
• Response viewer: inline floating panel to see AI response without switching apps
• Raycast / Alfred plugin parity