Setting up AI DevKit for Claude Code

Learn how to set up AI DevKit for Claude Code so your team can use the same operating model across AI coding agents.

AI DevKit adds project-local setup, memory, communication, workflow skills, and verification to Claude Code so every project follows the same playbook.

• Project initialization and environment selection
• AI editor files like rules, agents, and commands
• Memory, communication, verification, and skills for repeatable workflows AI DevKit gives your coding agents one operating layer for setup, supervision, communication, local-first memory, workflow skills, and verification.

Getting started has two scopes:

1. Once per machine: run setup to connect detected local agents, install their session integrations, and install AI DevKit's built-in skills globally.
2. Once per project: run init to create .ai-devkit.json, environment-specific project files, and workflow documentation.

Keeping these steps separate makes it clear which changes affect your machine and which files belong in your project.

Before you begin, make sure you have:

• Node.js 20.20.0 or newer

• npm or npx, which comes with Node.js

• tmux for interactive managed agents (provisional compatibility floor: tmux 2.6+; setup reports but does not reject older versions)

• At least one supported AI coding agent or environment

• Node.js 20.20.0 or newer

• npm or npx, which comes with Node.js

• At least one supported AI coding agent or environment

Install and launch your coding agent at least once before running setup. AI DevKit detects an agent from its home directory, so a newly installed agent that has never started may be reported as skipped.

Install the command, then set up detected agents:

npm install -g ai-devkit
ai-devkit setup

npm install -g ai-devkit installs the CLI but does not run setup for you.

In each project, run:

cd your-project
ai-devkit init

You can use AI DevKit without installing a global command:

npx ai-devkit@latest setup

Then initialize each project with npx too:

cd your-project
npx ai-devkit@latest init

An npx-only installation does not make a permanent ai-devkit command available. Prefix every later command with npx ai-devkit@latest, including agent, lint, memory, and skill commands.

setup checks for supported agent home directories. For each detected agent, it installs the available session hook or tracker and the AI DevKit built-in skills in that agent's global skill location.

Read the setup summary carefully. A skipped agent was not changed. If an agent you use is skipped, launch it once and rerun the same setup command.

Run init from the root of the project you want to use:

ai-devkit init

The interactive flow asks which project environments and workflow phases you want. It then:

1. Creates .ai-devkit.json with your project choices.
2. Creates environment-specific project templates.
3. Creates workflow documents under docs/ai/ by default.

If the directory is not already a Git repository and Git is available, init also initializes one.

Restart your coding agent after machine setup, then start an agent session in the initialized project. Check discovery before opening the console:

ai-devkit agent list

If the session appears, open the local console:

ai-devkit agent console

Then try sending a small task to the ID shown by agent list:

ai-devkit agent send "summarize the current branch and test status" --id <agent-id>

For an npx-only setup, use the same sequence with the npx prefix:

npx ai-devkit@latest agent list
npx ai-devkit@latest agent console
npx ai-devkit@latest agent send "summarize the current branch and test status" --id <agent-id>

CI normally needs project workflow artifacts, not local session hooks. Supply the project environment and phases explicitly, and use --built-in when CI needs project-local copies of the built-in skills:

npx -y ai-devkit@latest init \
  --yes \
  --environment <environment> \
  --all \
  --built-in

init --built-in is a CI and non-interactive convenience. For a normal local first run, use setup to install built-in skills globally.

For a repeatable team configuration, you can use an init template instead of listing every choice in the command.

The default workflow documentation path is docs/ai/. You can customize it during initialization or in .ai-devkit.json.

docs/ai/
├── requirements/    # What you're building and why
├── design/          # Architecture and technical decisions
├── planning/        # Task breakdown and timeline
├── implementation/  # Implementation notes and guides
├── testing/         # Test strategy and cases
├── deployment/      # Deployment procedures
└── monitoring/      # Monitoring and observability

These documents give agents durable context between phases instead of relying on chat history alone.

After agent list finds your session, ask your coding agent:

Use the dev-lifecycle skill to start requirements for a small feature.

The installed workflow skills guide requirements, design, planning, implementation, testing, verification, and review. See Development with AI DevKit for the full workflow.

1. Check environment-specific project support
2. Learn how project init and install work
3. Operate running agents
4. Give agents long-term memory
5. Manage skills
6. Install plugins

• If npx fails inside Codex, see Codex sandbox troubleshooting.
• Open an issue on GitHub for bugs or questions.

Once AI DevKit is set up in Claude Code, follow the getting started guide and ask the agent to use the dev-lifecycle skill so it plans before editing code.