iPhone’s Apple Intelligence shows a beautiful iridescent glow around the screen edges when Siri is thinking. I wanted that same effect on my Mac when Claude Code is working.
So I built EdgeGlow — a free, open-source macOS menu bar app that recreates iPhone’s edge glow effect on Mac.
In this post, I’ll walk through how I did it and the technical challenges I faced.
How It Works
AI Agent triggers hook → curl http://127.0.0.1:9876/start → Screen glows
AI Agent finishes → curl http://127.0.0.1:9876/stop → Glow fades out
Enter fullscreen mode Exit fullscreen mode
That’s it. Dead simple.
The interesting part is the animation. Let me walk you through it.
The Technical Challenge
I needed a smooth marquee effect around the screen edges — essentially a dashed line that flows continuously.
Why Not CABasicAnimation?
My first approach used CABasicAnimation on lineDashPhase:
let anim = CABasicAnimation(keyPath: "lineDashPhase")
anim.fromValue = 0
anim.toValue = perimeter
anim.duration = 5.0
anim.repeatCount = .infinity
shape.add(anim, forKey: "flow")
Enter fullscreen mode Exit fullscreen mode
This worked… until the window was hidden and shown again.
Core Animation would lose the animation state, and the marquee would:
- Freeze in place
- Reverse direction suddenly
- Jump to a random position
The Root Cause
CABasicAnimation relies on Core Animation’s animation system, which is tied to the layer’s presentation state. When the window is hidden (orderOut) or the layer is removed from the hierarchy, the animation state is lost.
I tried several workarounds:
- Pause/resume animations — didn’t work reliably
- Keep the window alive — wasted resources
- Reset animation on show — caused visual jumps
None of them were bulletproof.
The Solution: Timer-Driven Animation
I switched to a Timer at 60fps that directly updates lineDashPhase:
private var flowTimer: Timer?
private var dashPhase: CGFloat = 0
private var lastTickTime: CFTimeInterval = 0
private func startFlow() {
stopFlow()
lastTickTime = CACurrentMediaTime()
flowTimer = Timer.scheduledTimer(withTimeInterval: 1.0 / 60.0, repeats: true) { [weak self] _ in
self?.tickFlow()
}
RunLoop.main.add(flowTimer!, forMode: .common)
}
private func tickFlow() {
let now = CACurrentMediaTime()
let dt = CGFloat(now - lastTickTime)
lastTickTime = now
// Clamp dt to prevent huge jumps (e.g., returning from background)
let clampedDt = min(dt, 0.1)
let perimeter = currentPerimeter()
guard perimeter > 0, settings.animationDuration > 0 else { return }
let speed = perimeter / CGFloat(settings.animationDuration)
dashPhase += speed * clampedDt * (settings.clockwise ? 1 : -1)
ringLayer.sublayers?.forEach { layer in
(layer as? CAShapeLayer)?.lineDashPhase = dashPhase
}
}
Enter fullscreen mode Exit fullscreen mode
This is bulletproof — no dependency on Core Animation’s animation system, no state loss on window visibility changes.
The performance is also great: updating a single CGFloat property at 60fps uses ~0% CPU.
The 4-Layer + 20-Segment Glow Effect
To recreate iPhone’s Apple Intelligence edge glow, I needed two things:
1. 20-segment gradient coloring: Split the screen edge into 20 segments, each with a different hue (purple → blue → cyan → pink → orange → gold). Gaussian blur at segment boundaries creates smooth transitions:
for i in 0..<20 {
let hue = (baseHue + Double(i) * 0.05).truncatingRemainder(dividingBy: 1.0)
// Each segment has a different color, forming a gradient
}
Enter fullscreen mode Exit fullscreen mode
2. 4-layer glow stack: Stack 4 CAShapeLayer instances with different blur levels to create a realistic neon glow effect:
Layer Configuration
let configs: [(widthMul: CGFloat, alphaMul: CGFloat, blur: Double)] = [
(baseWidth * 1.5, 0.15, 12.0), // Layer 1: Wide line, high blur, low alpha → outer glow
(baseWidth * 0.8, 0.30, 8.0), // Layer 2: Medium line, medium blur, medium alpha → mid glow
(baseWidth * 0.3, 0.70, 2.0), // Layer 3: Thin line, low blur, high alpha → core line
(baseWidth * 0.1, 0.95, 0.0), // Layer 4: Thinnest line, no blur, full alpha → bright center
]
Enter fullscreen mode Exit fullscreen mode
Building Each Layer
for (lineWidth, alpha, blur) in configs {
let shape = CAShapeLayer()
shape.frame = ringLayer.bounds
shape.path = path
shape.fillColor = nil
shape.strokeColor = color.withAlphaComponent(alpha).cgColor
shape.lineWidth = lineWidth
shape.lineCap = .round
shape.lineDashPattern = [NSNumber(value: Double(dashLen)),
NSNumber(value: Double(gapLen))]
if blur > 0 {
let bf = CIFilter(name: "CIGaussianBlur")!
bf.setValue(blur, forKey: "inputRadius")
shape.filters = [bf]
}
ringLayer.addSublayer(shape)
}
Enter fullscreen mode Exit fullscreen mode
The Result
The 4 layers stack on top of each other, creating a realistic neon light tube effect:
Layer 4: ██ (bright center, no blur)
Layer 3: ████ (core line, blur 2)
Layer 2: ██████ (mid glow, blur 8)
Layer 1: ████████ (outer glow, blur 12)
Enter fullscreen mode Exit fullscreen mode
The outer layers create the soft glow halo, while the inner layers create the bright core. Together, they look like a real neon light.
Multi-Terminal Reference Counting
The Problem
If you have multiple Claude Code terminals running, one calling /stop would kill the glow even though others are still active.
The Solution: Reference Counting
class ControlServer {
private var activeCount = 0
private func processRequest(_ raw: String, conn: NWConnection) {
switch path {
case "/start":
activeCount += 1
onStart?()
resetSafetyTimer()
case "/stop":
activeCount = max(0, activeCount - 1)
if activeCount == 0 { onStop?() }
case "/pulse":
activeCount = max(0, activeCount - 1)
if activeCount == 0 { onPulse?() }
}
}
}
Enter fullscreen mode Exit fullscreen mode
Now /start increments the count, /stop decrements it. The glow only hides when the count reaches 0.
Safety Timeout
What if an agent crashes without sending /stop? The reference count would stay > 0 forever.
Solution: a 120-second safety timeout:
private func resetSafetyTimer() {
safetyTimer?.cancel()
let work = DispatchWorkItem { [weak self] in
guard let self = self else { return }
if self.activeCount > 0 {
print("⏰ 60s no activity, resetting to 0")
self.activeCount = 0
self.onStop?()
}
}
safetyTimer = work
DispatchQueue.main.asyncAfter(deadline: .now() + 60.0, execute: work)
}
Enter fullscreen mode Exit fullscreen mode
Every /start resets the timer. If no /start is received for 60 seconds, the count resets to 0.
Security Considerations
EdgeGlow runs an HTTP server on 127.0.0.1:9876. I had to be careful about security:
Localhost Only
let param = NWParameters.tcp
param.acceptLocalOnly = true // Only accept connections from localhost
Enter fullscreen mode Exit fullscreen mode
This prevents external network access.
GET Requests Only
guard method == "GET" || method == "OPTIONS" else {
sendResponse(405, "Method Not Allowed", conn: conn)
return
}
Enter fullscreen mode Exit fullscreen mode
Rejects POST/PUT/DELETE to prevent CSRF attacks.
No CORS Headers
// No Access-Control-Allow-Origin header
// Web JavaScript cannot invoke endpoints
Enter fullscreen mode Exit fullscreen mode
This prevents web pages from calling the API via JavaScript.
No Data Collection
- No analytics
- No telemetry
- No network requests (except localhost)
Multi-Monitor Support
The Challenge
How to handle multiple displays with different sizes?
The Solution
Calculate the union of all screen frames:
func totalScreenFrame() -> NSRect {
var frame = NSRect.zero
for screen in NSScreen.screens {
frame = NSUnionRect(frame, screen.frame)
}
return frame
}
Enter fullscreen mode Exit fullscreen mode
Handling Display Changes
Listen to screen parameter changes:
NotificationCenter.default.addObserver(
forName: NSApplication.didChangeScreenParametersNotification,
object: nil,
queue: .main
) { [weak self] _ in
// Debounce 500ms to avoid rapid rebuilds
self?.screenChangeWorkItem?.cancel()
let work = DispatchWorkItem { [weak self] in
self?.handleScreenChange()
}
self?.screenChangeWorkItem = work
DispatchQueue.main.asyncAfter(deadline: .now() + 0.5, execute: work)
}
Enter fullscreen mode Exit fullscreen mode
When displays change, rebuild the layers with the new frame size.
Settings Persistence
The Challenge
How to make settings reactive and persist them across launches?
The Solution
Use ObservableObject + UserDefaults + Combine:
class AppSettings: ObservableObject {
@Published var speed: Int {
didSet {
UserDefaults.standard.set(speed, forKey: "speed")
notifyChange()
}
}
static let shared = AppSettings()
private init() {
self.speed = UserDefaults.standard.integer(forKey: "speed")
if speed == 0 { speed = 5 } // default
}
}
Enter fullscreen mode Exit fullscreen mode
Then in GlowWindow, observe changes:
settings.$speed.sink { [weak self] _ in
self?.rebuildLayers()
}.store(in: &cancellables)
Enter fullscreen mode Exit fullscreen mode
When the user changes the speed slider, the layers rebuild automatically.
Performance
Metric Value Notes CPU ~0% Timer 60fps only updates one CGFloat Memory ~50MB 4 CAShapeLayer + CIFilter Disk 892KB Universal Binary (arm64 + x86_64) Network 0 localhost only, no external requestsThe key insight: updating a property at 60fps is cheap. It’s the GPU-accelerated CAShapeLayer rendering that does the heavy lifting.
Try It
- GitHub: https://github.com/vector4wang/EdgeGlow
- Download: https://github.com/vector4wang/EdgeGlow/releases
- Size: 892KB zipped
- License: MIT
- macOS: 13.0+ (Ventura)
It’s my first macOS app. Built with pure Swift + SwiftUI, no third-party dependencies.
Would love to hear your feedback!
답글 남기기