Uatu: AI Orchestration Framework and Agent Collaboration System for Claude Code

Uatu is an AI orchestration framework designed exclusively for Claude Code, aiming to address the challenges of workflow management for AI agents in AI-assisted programming. Through a three-layer architecture of rules, hooks, and commands, it provides capabilities for proactive suggestions, multi-agent execution, specification-driven development, and prompt quality coaching. Its core value lies in elevating AI assistance from a single function to systematic workflow orchestration, helping developers efficiently manage AI agent collaboration.

With the popularization of AI-assisted programming tools, effectively managing and orchestrating the workflows of AI agents has become a new challenge for developers. As an AI orchestration framework, Uatu can be installed once and used in any project without manual editing of the CLAUDE.md file. The framework name "Uatu" is derived from The Watcher in Marvel Comics, symbolizing its ability to continuously observe project status and provide intelligent suggestions.

Each layer of Uatu's three-layer architecture has its own responsibilities:

1. Rule Layer: Located at .claude/rules/uatu-core.md, it is automatically loaded in each session to ensure the consistency and predictability of AI behavior, guiding Claude to proactively suggest commands, coach prompt quality, etc.
2. Hook Layer: Triggered during Claude Code lifecycle events to implement automated workflows (such as prompt scoring, agent suggestions, test missing warnings, branch protection, etc.).
3. Command Layer: Provides 7 slash commands for users to explicitly control AI behavior, complementing the automated hooks.

Uatu's core commands include:

• /status: Runs at the start of a session to provide the complete status of the project (Sprint progress, branch status, worktree, etc.), with the --all parameter supporting multi-project viewing.
• /orchestrate: Automatically analyzes tasks and selects the appropriate workflow (e.g., bug fixes trigger a 4-stage debugging process, refactoring triggers an architecture review process), supporting combinable flags like --tdd, --e2e, --jira.
• Other commands: /research (in-depth research), /debug (structured debugging), /spec (specification-driven development), /review (intelligent code review), /ship (one-click delivery).

Steps for installing and configuring Uatu:

• New machine: Clone the repository → run setup.sh → install the user-level MCP server (uatu-setup).
• New project: uatu-install (supports configuration files like standard, minimal, full, and allows specifying the language).
• Existing project: Also use uatu-install, which does not overwrite the existing CLAUDE.md and automatically adds rules and hooks.

Practical application scenarios of Uatu:

1. New feature development: After describing the requirements, it automatically triggers the process of research → planning → wave execution → review, and performs parallel architecture analysis and solution generation.
2. Production bug fix: Identifies bug keywords and triggers a 4-stage debugging process: capture → isolation → fix → verification.
3. Code refactoring: Identifies refactoring keywords and triggers the process of architecture review → parallel refactoring → testing → review.

Uatu's design philosophy is reflected in:

• Automation first: Reduce manual intervention through hooks, allowing developers to focus on core logic.
• Intelligent routing: The /orchestrate command automatically selects the best workflow, eliminating decision fatigue.
• Composability: Flag combinations adapt to the needs of different scenarios.
• Progressive adoption: From minimal installation to full functionality, supporting on-demand configuration.
• Ecosystem integration: Seamlessly integrates with tools like Jira and GitHub, fitting into existing development processes.

Uatu represents the evolutionary direction of AI-assisted development tools—from single-function code completion to systematic workflow orchestration. Through its three-layer architecture, it brings enterprise-level AI orchestration capabilities to Claude Code. As AI plays an increasingly important role in software development, such orchestration frameworks will become standard tools for development teams.