Recap is an open-source, macOS-native application designed to help users summarize meetings without compromising privacy. Built for professionals who need to capture key insights from audio recordings but don’t want their sensitive conversations uploaded to cloud servers, Recap leverages Apple’s Core Audio and Swift frameworks to record system audio and microphone input directly on the device. It then transcribes speech using WhisperKit (MLX) and generates summaries via Ollama or OpenRouter — with all processing occurring locally unless explicitly configured otherwise. The app targets developers, remote workers, and privacy-conscious teams who rely on tools like Zoom, Microsoft Teams, or Google Meet but need a secure way to review meeting content without external data exposure.
Currently in early development, Recap is a proof-of-concept that prioritizes transparency and local processing over convenience. While not yet production-ready, it provides a functional foundation for users who want full control over their audio data and are willing to compile from source. The project is actively seeking contributors to stabilize features like meeting detection, live transcription, and background processing.
What You Get
- Meeting Detection - Automatically detects audio from popular meeting apps like Microsoft Teams, Zoom, and Google Meet using macOS ScreenCaptureKit (in development).
- Audio Recording - Captures system audio and optionally blends in microphone input for dual-source recording using native Core Audio taps and AVAudioEngine.
- Local Processing - Uses WhisperKit (MLX) for on-device transcription and Ollama for local LLM-based summarization, ensuring no audio or transcripts leave your Mac.
- Privacy First - All data is processed locally; cloud options (OpenRouter) are optional and require explicit API key configuration.
- Open Source - Full codebase is publicly available on GitHub with MIT licensing, enabling community review and contributions.
Common Use Cases
- Building a private meeting archive for compliance - Legal or finance teams recording internal strategy calls and needing auditable, on-device transcripts without cloud storage risks.
- Remote developers managing back-to-back syncs - Engineers who want to review meeting highlights without switching contexts, using Recap to auto-summarize Zoom calls while coding.
- Problem: Sensitive discussions in cloud-based tools → Solution: Recap processes audio locally - Teams discussing product roadmaps or client contracts avoid exposing transcripts to third-party AI services by using Ollama-hosted models on their M-series Macs.
- Team workflow for non-administrators - Contributors in organizations with strict data policies use Recap to generate summaries without IT approval, since no network calls are made by default.
Under The Hood
RecapAI is a macOS menu bar application designed to automate the recording and transcription of audio from meetings, integrating system-level audio capture with AI-powered speech recognition and intelligent app detection. It leverages modern Swift frameworks and Apple’s ecosystem tools to deliver a seamless user experience for meeting management and summarization.
Architecture
The application adopts a layered architecture that clearly separates concerns across audio processing, data handling, and UI components. It uses dependency injection and factory patterns to manage object creation and lifecycle.
- The architecture follows a modular structure with distinct layers for audio capture, processing, file management, and UI interaction
- Design patterns such as coordinator, service locator, and delegate are employed to manage complex workflows and UI state
- Component interactions are governed through type-safe protocols and dependency containers that promote loose coupling and maintainability
- Module organization emphasizes separation of concerns with clear boundaries between audio, data, and presentation layers
Tech Stack
Built as a Swift-based macOS application, RecapAI integrates deeply with Apple’s ecosystem to enable robust audio and screen recording capabilities.
- The application is developed entirely in Swift, utilizing SwiftUI for UI and AppKit for menu bar integration, with AVFoundation and ScreenCaptureKit for core audio functions
- Key dependencies include WhisperKit for transcription, CoreData for local data persistence, and Toast for user notifications
- It is built using Xcode with standard Apple frameworks such as Foundation, Combine, and OSLog for system-level integration
- Testing is supported through XCTest and Mockable, enabling unit and UI testing with mockable dependencies for improved testability
Code Quality
The codebase reflects a moderate level of quality with some structured organization and clear separation of concerns, although it lacks comprehensive test coverage and exhibits inconsistent patterns.
- Testing efforts are minimal with no identifiable test files, indicating a lack of automated validation and reliability checks
- Error handling is basic, relying on standard try/catch mechanisms without centralized or advanced error propagation systems
- Code consistency varies with some adherence to Swift conventions, though naming and structure could be more uniform across modules
- Technical debt is visible in duplicated logic, limited abstraction, and unclear module boundaries that hinder long-term maintainability
What Makes It Unique
RecapAI distinguishes itself through its integration of low-level audio capture with high-level AI transcription and intelligent meeting detection.
- It combines system-level audio input using ScreenCaptureKit and AVFoundation with WhisperKit for real-time speech-to-text transcription and meeting summarization
- A modular dependency injection system cleanly separates concerns between audio capture, processing, and data management components
- The architecture supports extensibility through service-oriented modules that can be swapped or extended without affecting core functionality
- It enables intelligent app detection and automatic meeting recognition, offering a unique blend of automation and AI-driven insights