Yuma Uchida

personal context - Local, portable, + open source context across all LLMs

Tired of rewriting your background for every new AI ? LLM memory features are great, but your context gets trapped inside each platform. Personal Context Engine solves this. It passively interviews you to extract "vignettes" (behavioral scenes, not just labels) and exports your true identity into portable Markdown files. Use the built-in MCP server to connect your local context directly to Claude Desktop or Cursor. No cloud lock-in. Stop the generic AI replies. Own your context.

Add a comment

Replies

Best
Yuma Uchida
Hi Product Hunt! 👋 I’m Yuma, the creator of Personal Context Engine. A bit about my journey: I recently lost my business, fell into debt, and left Tokyo with just one suitcase and a laptop to rebuild my life from absolute zero. 🧳💻 Naturally, I turned to AI (ChatGPT, Claude) for advice on how to start over and find the right path for me. But the answers were always terrible. I would ask for business ideas, and it would suggest generic things like "start dropshipping" or "do affiliate marketing"—things that completely ignored who I actually am. I realized the problem wasn't the AI's intelligence. The problem was that the AI didn't have my true context. Even if I pasted a brief profile into the system prompt, it didn't help. Because what we think our values are (e.g., "I value freedom") is often just a filtered, socialized version of ourselves. I realized I needed the AI to know my actions and sensations—the specific prerequisites of my life. So, I built an open-source engine to fix this: 1. Passive Interview: It asks you behavior-based questions (e.g., "Where did you actually spend your time and money this week?" or "What made your body feel light?"). 2. Extracts "Vignettes": It captures specific scenes of your life, not just boring bullet points. 3. Portable Markdown: It exports your true identity into plain .md files. 4. Built-in MCP Server: Connect your local context directly to Claude Desktop or Cursor seamlessly. No cloud lock-in. It's entirely Open Source (MIT) and runs locally. I’d love for you to try the live demo (no install required) and let me know what you think about this approach to making AI actually understand you! GitHub: [https://github.com/uchidayuma/pe...