Xavier

Xavier is a runtime-agnostic AI agent orchestrator that manages a personal knowledge vault, coordinates parallel reviewer agents, and gets smarter with every interaction.

Think of it as a global dev tool — like Homebrew or Oh My Zsh — that ties your AI coding skills together, remembers what it learns, and adapts to how you work.

Get started

curl -fsSL https://xavier.team/install | sh

What's new in 0.4.0 — new /xavier investigate skill for hypothesis-driven bug investigation, scoped mode for /xavier learn [path] in monorepos, and a new investigation vault type. Read the changelog.

What it does

• Panel code reviews — multiple persona-driven reviewers analyze your changes in parallel
• Knowledge vault — a Zettelkasten-based store that compounds insights across repos, teams, and sessions
• Dependency skills — best-practice summaries for your key packages, always up to date
• PRD creation — interview-driven product specs enriched with vault context
• Task decomposition — break PRDs into phased vertical slices with acceptance criteria
• Autonomous execution — an iteration loop that executes task phases with built-in guardrails
• Codebase learning — scan a repo and generate architecture maps, decision logs, and dependency graphs
• Hypothesis-driven investigation — parallel remoras chase root causes for bugs and surface ranked diagnoses
• Topic-driven research — fan out across web, internal docs, and your codebase to produce structured digests on any concept
• Obsidian bridge — export vault notes to your personal knowledge base

Or jump into the guide to understand what Xavier brings to your workflow.