Introduction

Xavier is an AI agent orchestrator and knowledge system. It's installed globally on your machine, manages a personal knowledge vault, and coordinates multiple AI agents to help you review code, plan features, and execute tasks — all informed by what it's learned from your past work.

The problem

AI coding tools are powerful individually, but they don't talk to each other. A PRD written in one session is invisible during a code review in another. Dependency documentation gets re-fetched every time. Architectural context about your codebase vanishes when the conversation ends.

There's no system that ties these capabilities together, learns over time, and adapts to how you actually work.

The solution

Xavier sits between you and your AI coding agents. It provides:

• A unified knowledge vault (~/.xavier/) that persists across sessions, repos, and tools
• Concurrent agent orchestration using the Shark pattern for non-blocking, parallel execution
• Persona-driven review panels where multiple agents analyze your code from different angles
• Knowledge that compounds — every review, scan, and decision is stored and informs future operations

Not a plugin

Xavier is distributed as a standalone tool via curl | sh, not as a plugin for any specific AI runtime. The knowledge layer (vault, personas, config) is plain files. The execution layer uses pluggable runtime adapters. Claude Code is the primary adapter, but the architecture supports others.

Prerequisites

You need three things installed:

• Git — Xavier's vault is a git repository
• gh CLI — for creating and syncing your vault's remote repository
• A compatible AI coding runtime (Claude Code, etc.)

Next steps

Ready to try it? Head to the Quick Start.

Want to understand the architecture? Check How It Works.