MCP Memory Service

# MCP Memory Service - Demo Recording Script **Duration Target:** 30 seconds max **Output:** High-quality screen recording for GIF conversion **Goal:** Show dramatic before/after difference --- ## 🎬 Recording Setup ### Before You Start 1. **Close unnecessary apps** (clean desktop) 2. **Set screen resolution** to 1920x1080 or 1280x720 3. **Disable notifications** (Do Not Disturb mode) 4. **Prepare two Claude Desktop sessions:** - Session A: Fresh session WITHOUT memory context - Session B: Fresh session WITH memory enabled 5. **Pre-populate memory** with project context: ```bash # Use /memory-store to add context before recording claude /memory-store "Project: Next.js app with App Router, Prisma ORM, tRPC for API, TypeScript, Tailwind CSS" claude /memory-store "Architecture decision: Using server components for data fetching, client components only when needed" claude /memory-store "Current focus: Building authentication system with NextAuth.js" ``` ### Recording Tools **macOS (Recommended):** ```bash # Option 1: QuickTime Player # File → New Screen Recording → Click red record button → Select area # Option 2: Built-in Screenshot.app # Cmd+Shift+5 → Record Selected Portion → Select Claude window → Record # Option 3: ffmpeg (for automation) screencapture -v demo-raw.mov ``` **Windows:** - Windows + G (Game Bar) - OBS Studio - ScreenToGif --- ## 🎭 Scene 1: WITHOUT Memory (10 seconds) **Setup:** Fresh Claude Desktop session, memory service disabled ### Actions: 1. **[0:00-0:02]** Show empty Claude chat - Visible: "New conversation" at top 2. **[0:02-0:05]** Type and send: ``` What tech stack am I using for my current project? ``` 3. **[0:05-0:10]** Claude responds (expected): ``` I don't have any information about your current project's tech stack. Could you tell me what technologies you're using? ``` 4. **[0:10]** **Visual indicator:** Red ❌ or "Context Lost" overlay --- ## 🎭 Scene 2: WITH Memory (20 seconds) **Setup:** Fresh Claude Desktop session, memory service enabled & populated ### Actions: 1. **[0:10-0:12]** Show NEW empty Claude chat (different session) - Visible: "New conversation" at top - **Optional:** Show memory service running indicator 2. **[0:12-0:15]** Type and send (SAME QUESTION): ``` What tech stack am I using for my current project? ``` 3. **[0:15-0:22]** Claude responds (expected): ``` Based on our previous work, you're building a Next.js application with the App Router, using Prisma for the ORM, tRPC for type-safe APIs, TypeScript, and Tailwind CSS for styling. You're currently focused on implementing authentication with NextAuth.js. ``` 4. **[0:22]** **Visual indicator:** Green ✓ or "Context Remembered" overlay 5. **[0:22-0:25]** Type follow-up: ``` Add OAuth login with Google ``` 6. **[0:25-0:30]** Claude responds (expected): ``` I'll integrate Google OAuth with your existing NextAuth.js setup in your Next.js App Router. Let me add the Google provider... ``` 7. **[0:30]** **End frame:** Show MCP Memory Service logo + "Never explain twice" --- ## 🎨 Visual Enhancements (Optional) Add these in post-processing or during recording: 1. **Text overlays:** - "WITHOUT Memory" (Scene 1) - "WITH Memory" (Scene 2) - "Same question, different results" 2. **Indicators:** - Red X for "Context Lost" - Green checkmark for "Context Remembered" 3. **Speed adjustments:** - Speed up typing (1.5x) - Slow down Claude responses for readability 4. **Highlight boxes:** - Around key phrases: "I don't have information" vs "Based on our previous work" --- ## 📹 Recording Best Practices ### Do: ✅ Record at **1920x1080** or **1280x720** (GIF-friendly) ✅ Use **60 FPS** if possible (smoother GIF) ✅ **Zoom in** on Claude window (fill screen) ✅ Use **dark mode** (better contrast, smaller file size) ✅ Keep mouse movements **smooth and deliberate** ✅ **Pause 2 seconds** before each action ### Don't: ❌ Show your full desktop (focus on Claude window only) ❌ Move mouse erratically ❌ Record at 4K (GIF will be huge) ❌ Include audio (GIFs are silent) ❌ Exceed 30 seconds (attention span limit) --- ## 🎞️ Post-Recording Checklist After recording, save the file as `demo-raw.mov` or `demo-raw.mp4` in the repo root, then: 1. **Trim** to exactly 30 seconds 2. **Add text overlays** (optional) 3. **Convert to optimized GIF** (I'll help with this) 4. **Upload to GitHub repo** under `docs/assets/demo.gif` 5. **Update README.md** with GIF link --- ## 🚀 Ready to Record? ### Quick Start Commands ```bash # macOS: Start recording with 3-second countdown sleep 3 && screencapture -v ~/Documents/GitHub/mcp-memory-service/demo-raw.mov # Then follow the script above, staying within 30 seconds ``` **When you're done recording, let me know and I'll help convert it to an optimized GIF!** --- ## 📝 Alternative: Screen Recording Services If you prefer professional polish, consider: 1. **Loom** (loom.com) - Easy recording + built-in GIF export 2. **ScreenFlow** (macOS) - Professional editing tools 3. **Camtasia** - Cross-platform with annotations These can add overlays, text, and export optimized GIFs automatically. --- **Questions?** Just ask! I'll help with any step of the process.