Welcome to PAI

Your Personal AI Infrastructure is ready. Here's how to get started and learn the system.

Your Portal

Your assistant creates visual content here. Bookmark these URLs on your host machine.

Portal (web content)
Loading...
File Exchange (drag and drop files)
Loading...
Clipboard (copyable text)
Loading...

Things to Try

Open Claude Code in the terminal and try these prompts. Your assistant will create something and give you a URL to view the result.

Explore PAI

Click through these pages to understand what your assistant is made of. Every skill and agent definition is readable — PAI is open by design.

⚡
Skills Catalog
Browse all skills. Click any skill to read its full definition and see how it's structured.
🤖
Agent Roster
Meet the specialized agents. Each has a unique personality, expertise, and backstory.
⚙
System Overview
See your installation at a glance: skill count, agents, Algorithm version, and directory layout.

Want to create your own skill? Open the Skills Catalog and click on CreateSkill. That's the skill that teaches your assistant how to build new skills. You can ask your assistant: "Create a skill for [whatever you want]" and it will follow that blueprint.

How PAI Works

PAI is built on a few core concepts. Understanding these will help you get more out of your assistant.

The Algorithm
Every non-trivial request goes through a 7-phase problem-solving process. Your assistant observes what you need, thinks about the best approach, plans the work, builds it, executes it, verifies it worked, and learns from the result. This is why it's systematic rather than just winging it.
1. Observe 2. Think 3. Plan 4. Build 5. Execute 6. Verify 7. Learn
Skills
Skills are modular capabilities. Each skill lives in its own folder with a SKILL.md file that defines what it does and when it activates. Skills can have workflows (multi-step procedures) and tools (scripts). Your assistant loads the right skills automatically based on what you ask. Browse them all.
Agents
When a task is complex, your assistant can delegate work to specialized agents. An Engineer agent builds code. An Architect designs systems. A Researcher investigates questions. Multiple agents can work in parallel. See the full roster.
Hooks
Hooks are automation triggers that run at specific moments — when a session starts, before a tool is used, after a response is generated. They handle things like formatting, capability detection, and status tracking without you having to think about them.
Ideal State Criteria
The core innovation of the Algorithm. For every task, your assistant creates a set of testable criteria that define what "done" looks like. These start as observations about what you want, evolve during planning, and become the actual verification checklist. If all criteria pass, the task is complete. This is what makes it reliably good rather than occasionally good.

How It Works Day-to-Day

Portal pages: When your assistant creates something visual (a report, dashboard, or styled page), it writes it to ~/portal/ and gives you a URL. Open it in your browser.

File exchange: Need to give your assistant a file? Drag it into the Exchange page. Need a file from your assistant? It will put it in the exchange and tell you.

Clipboard: When your assistant writes text you need to copy (an email, a message, a code snippet), it puts it on the Clipboard page with a one-click copy button.

Tips

Be specific. "Make me a budget spreadsheet with columns for date, description, amount, and category" works better than "help with my budget."

Iterate. Your assistant learns from the conversation. If something isn't right, tell it what to change. It keeps the full context.

Data stays local. Everything on this VM is only accessible from your local network. Nothing is shared publicly unless you specifically ask.

Explore the source. Every skill, agent, and hook is a readable text file. Ask your assistant to show you any of them. Understanding the system helps you use it better.