Getting Started

H·AI·K·U works with both claude.ai and Claude Code. Choose the installation method that fits your team.

1. Install the Plugin

claude.ai (Recommended for Teams)

Install H·AI·K·U as a claude.ai plugin for your organization or individual projects.

Organization-wide

Download the plugin zip and upload it through your organization's admin settings. This makes H·AI·K·U available to every project and team member.

Download Plugin (.zip)

Internal use via GitHub

Add the GitHub repository URL to your organization's plugin configuration to install directly:

https://github.com/TheBushidoCollective/haiku-method.git

Claude Code (CLI)

Point Claude Code at the GitHub repository or install via the CLI:

# Install from the marketplace
/install haikumethod.ai/marketplace.json
# Or point directly at the repo
/install TheBushidoCollective/haiku-method

2. Run Setup

The setup skill configures your workspace, discovers organizational context, and writes your settings. Run it once per project:

# Configure workspace and organizational context
/setup

Setup walks you through choosing a workspace location, discovering your team's tools and conventions, selecting a default workflow, and configuring quality gates. It's idempotent — re-run anytime to update settings.

3. Start Your First Intent

With the plugin installed and workspace configured, elaborate your first intent:

# Elaborate your first intent
/elaborate

The Elaboration skill guides you through defining your intent, breaking it into units, setting success criteria, and building a domain model. When you're ready, use /execute to begin structured execution with hat-based workflows.

Plugin Skills

/setup

Configure workspace, discover organizational context, set default workflow and quality gates.

/elaborate

Define intent, decompose into units, set success criteria, build domain model.

/execute

Execute work through hat-based workflows with iterative bolts and quality gates.

/operate

Manage delivered outcomes with recurring tasks, reactive responses, and AI guidance.

/reflect

Analyze outcomes, capture learnings, feed forward into organizational memory.

/resume

Resume work on an existing intent from where you left off.

/advance

Advance a unit to its next hat in the workflow.

For Software Teams

If you're building software, install the AI-DLC profile — it extends H·AI·K·U with git integration, test suites, PR workflows, CI/CD pipelines, and deployment gates.

# Install AI-DLC from the H·AI·K·U marketplace
/install ai-dlc

AI-DLC provides software-specific hats, workflows, and quality gates pre-configured for iterative development.

For Any Domain

H·AI·K·U works for any domain — marketing, operations, research, strategy. The core plugin gives you the full methodology. Here's the lifecycle:

Elaboration

Define intent, decompose into units, set success criteria. This alone transforms how you work with AI.

Execution

Work through hat-based workflows in iterative bolts. Quality gates ensure standards before progressing.

Operation

Manage delivered outcomes with recurring tasks, reactive responses, and AI-guided manual activities.

Reflection

Analyze outcomes, capture learnings, feed forward into organizational memory. This is where H·AI·K·U's compounding advantage begins.

Choosing a Collaboration Mode

The right mode depends on context. Use this as a starting guide:

Use Supervised when...

The work is novel, high-risk, or foundational. You're exploring unfamiliar territory. Mistakes would be costly to reverse.

Use Observed when...

The work is somewhat familiar but requires judgment calls. You want to see how AI approaches the problem. You're building trust in the process.

Use Autonomous when...

The work is well-defined with verifiable success criteria. Quality gates can enforce standards. You trust the process and want efficiency.