CaptureFlow

CaptureFlow is an open-source iOS app and reference architecture for turning screenshots into structured AI insight cards.

Summary

CaptureFlow turns screenshots and captured images into actionable AI insight cards, helping users prevent useful screenshots from becoming forgotten camera-roll clutter.

The app can import or capture an image, analyze it with an AI vision provider, generate an editable card, save it locally, export it as Markdown, and turn supported cards into Reminders or Calendar events.

The current AI-backed provider uses OpenAI for screenshot/image analysis and generation. Apple Foundation Models can be used for card generation only on supported iOS 26+ devices.

The codebase is designed with clear service boundaries, so developers can replace the default LLM, prompt, storage, or action integrations.

Why CaptureFlow?

Most screenshots contain useful context, but they usually become forgotten camera-roll clutter. CaptureFlow turns screenshots into structured, editable, and reusable insight cards.

Unlike a one-off chat prompt, CaptureFlow provides a reusable iOS pipeline for:

• vision analysis
• prompt-based card generation
• local persistence
• Markdown export
• Reminder and Calendar integration
• provider-based LLM replacement

Features

• Capture or import screenshots and images
• Analyze visual context with an AI vision provider
• Generate editable AI insight cards
• Save cards locally on device
• Export cards as Markdown
• Create Reminders and Calendar events from supported cards
• Bring your own LLM provider
• Optional Apple Foundation Models generation on iOS 26+
• Designed as a reference architecture for AI-native iOS apps

Quick Start

1. Clone the repository and open CaptureFlow.xcodeproj in Xcode.
2. Configure an OpenAI API key for AI-backed screenshot analysis and generation.
3. Build and run the CaptureFlow scheme on an iOS simulator or device.
4. Optional: switch generation between External LLM and Apple Foundation Models in Settings when Foundation Models are available.

An OpenAI API key is required for AI-backed screenshot/image analysis. Without a key, the app can still be built, but AI analysis features will not produce real results.

API Key Setup

API keys are not entered in the app UI. Configure OpenAI-backed analysis and generation with Xcode scheme environment variables:

CAPTUREFLOW_AI_PROVIDER=openai
CAPTUREFLOW_OPENAI_API_KEY=your-api-key

OPENAI_API_KEY is also accepted as a fallback key name. These values can also be supplied in CaptureFlow/Resources/LocalSecrets.plist, which is ignored by git:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
  <key>CAPTUREFLOW_AI_PROVIDER</key>
  <string>openai</string>
  <key>CAPTUREFLOW_OPENAI_API_KEY</key>
  <string>your-api-key</string>
</dict>
</plist>

Foundation Models Support

Apple Foundation Models are used for card generation only when available. Screenshot/image vision analysis currently still requires an external vision-capable provider such as OpenAI.

Foundation Models support requires:

• iOS 26+
• supported Apple Intelligence devices
• availability of Apple Foundation Models on the device

Users can switch generation between External LLM and Apple Foundation Models from Settings when Foundation Models are available.

Bring Your Own Provider

CaptureFlow is provider-based by design. You can replace the default OpenAI implementation with your own provider by implementing:

• VisionAnalysisProviding for screenshot/image understanding
• LLMProviding for card generation

Then register your implementation in AppContainer.local().

• Use LLMProviderConfiguration to define provider display name and default model names.
• Replace VisionAnalysisPromptProviding or CardGenerationPromptProviding to customize prompts.
• Features should continue to depend on VisionAnalyzing and CardGenerating, not concrete provider implementations.

For more detail, see Architecture.

Build & Test

xcodebuild -project CaptureFlow.xcodeproj -scheme CaptureFlow -destination 'generic/platform=iOS Simulator' IPHONEOS_DEPLOYMENT_TARGET=18.0 build

Run tests:

xcodebuild test -project CaptureFlow.xcodeproj -scheme CaptureFlow -destination 'platform=iOS Simulator,name=iPhone 16 Pro Max,OS=18.6'

If your simulator name differs, replace the destination with an installed simulator.

Requirements

• iOS 18.0 minimum deployment target
• Xcode 26.3 or newer is recommended
• iOS 18 simulator runtime or newer
• Apple Foundation Models require iOS 26+ and supported devices

Roadmap

Near-term

• OCR-only baseline mode for text extraction and basic note cards without an external vision provider
• User-facing prompt presets for tone, detail level, and preferred card style
• More export formats, starting with plain text, JSON, and iOS share sheet support
• Regenerate or rewrite individual generated sections
• Better failure recovery, including retrying analysis or generation separately

Card and action expansion

• More card templates, starting with receipt, contact, article, product, and travel cards
• More concrete actions, such as create contact, open detected URL, create checklist reminder, and share card
• Tags, collections, and stronger inbox organization for long-term saved card management

Integrations

• Share Extension for sending screenshots and images into CaptureFlow from Photos, Safari, and other apps
• App Intents and Shortcuts, such as analyzing the latest screenshot or creating a reminder from a captured image
• API key and provider settings UI backed by Keychain storage

Architecture-heavy milestones

• Multiple screenshot support across capture, analysis, persistence, and saved source metadata
• True partial generation support with streamed section-level generation events
• macOS or Mac Catalyst support focused on drag-and-drop image input, analysis, saved cards, and supported actions

Open Source

• Contributions: CONTRIBUTING.md
• Code of conduct: CODE_OF_CONDUCT.md
• Security policy: SECURITY.md
• License: MIT