v0.11.0

MemoryLane v0.11.0

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Native Swift screenshot module — added a new native macOS screenshot capture path using Swift, replacing the Electron desktopCapturer dependency for faster, lower-overhead screen capture
• Fixed-interval ScreenCapturer — new ScreenCapturer class that supports fixed-interval frame capture alongside the existing event-driven mode
• Configurable capture settings UI — adjust visual change threshold, typing timeout, and scroll timeout directly from the main window (#28, closes #33)
• URL parsing fix — fixed URL parsing in the bundled app that could cause failures in production builds
• Internal improvements — better bundleId separation for dev/production builds, updated MCP prompt quality, increased max history size, added migration tooling and architecture documentation

Features

• Automatic updates — background update checks with one-click install from the tray menu
• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models — use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing — groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

v0.10.0...v0.11.0