How to Build a Personal Context Portfolio and MCP Server

What It Covers

The AI Breakdown presents a framework for building a Personal Context Portfolio — 10 structured markdown files that serve as a portable, machine-readable operating manual for any AI agent or tool, eliminating the repetitive tax of re-explaining yourself every time you onboard a new AI system.

Key Questions Answered

• •Context Repetition Tax: Every new AI agent setup requires re-explaining your role, projects, preferences, and constraints from scratch. As agent usage scales from 3 to 10+ tools weekly, this tax degrades output quality — not just time — because incomplete context explanations leave critical information out, producing generic rather than personalized results.
• •10-File Portfolio Structure: Build your context portfolio across 10 markdown files: identity, roles and responsibilities, current projects, team and relationships, tools and systems, communication style, goals and priorities, preferences and constraints, domain knowledge, and decision log. Modular design lets individual agents pull only relevant files rather than processing one monolithic document.
• •Decision Log as Underrated Asset: The decision log file — recording past choices and their reasoning — provides agents with behavioral precedent when helping evaluate new decisions. This transforms agents from generic advisors into systems that understand your specific decision-making patterns, making recommendations consistent with your established judgment history.
• •AI-Interviewed Portfolio Creation: Build portfolio files using an interview loop rather than writing manually. Create a Claude or ChatGPT project, then run each file through an interview-to-draft-to-revision cycle. A companion app built on Claude Opus handles all 10 files simultaneously, routing single answers across multiple relevant files and offering free portfolio download.
• •Local-to-Remote MCP Deployment: Convert your markdown portfolio into an MCP server by working step-by-step with an AI build partner. Start locally, then push to GitHub and deploy via Railway for remote access. Most time is spent troubleshooting — request complete code blocks rather than partial edits to avoid copy-paste errors that cause the majority of setup failures.