How to Install Arc-Kit: Complete Setup Guide for AI Assistants

Arc-Kit installs as a Python package via marketplace plugins, CLI extensions, or pip depending on your AI assistant platform, providing 68 slash commands and 10 autonomous agents.

Arc-Kit is an open-source architecture toolkit distributed by tractorjuice/arc-kit that enhances AI coding assistants with enterprise architecture capabilities. Whether you use Claude Code, Gemini CLI, GitHub Copilot, or Codex CLI, installing Arc-Kit equips your environment with specialized research agents, MCP servers, and architecture generation commands.

Prerequisites

Before you install Arc-Kit, ensure your environment meets the version requirements defined in pyproject.toml【/blob/main/pyproject.toml#L5-L6】:

• Python 3.11 or higher
• pip or uv package manager
• Compatible AI assistant platform (Claude Code v2.1.112+, Gemini CLI, GitHub Copilot, or Codex/OpenCode CLI)

Install Arc-Kit for Claude Code

Claude Code is the primary development platform for Arc-Kit. The installation uses the marketplace plugin system described in the repository README【/blob/main/README.md#L38-L44】.

1. Add the marketplace plugin:

/plugin marketplace add tractorjuice/arc-kit

2. Enable the plugin from the Claude Code Discover tab.

3. Verify installation – The plugin automatically registers all 68 slash commands, 10 autonomous research agents, and bundled MCP servers (AWS Knowledge, Microsoft Learn, Google Developer Knowledge, DataCommons, and govreposcrape).

Updates are automatic when the marketplace repository changes.

Install Arc-Kit for Gemini CLI

For Gemini CLI users, Arc-Kit installs as an extension using the command referenced in README lines 48-52【/blob/main/README.md#L48-L52】:

gemini extensions install https://github.com/tractorjuice/arckit-gemini

This copies the command set, templates, and MCP servers into the Gemini extension directory. Update future versions with:

gemini extensions update arckit

Install Arc-Kit for GitHub Copilot

GitHub Copilot users install Arc-Kit via Python pip and scaffold a project using the arckit init command. The installation steps follow README lines 58-64【/blob/main/README.md#L58-L64】:

# Install the core package from source

pip install git+https://github.com/tractorjuice/arc-kit.git

# Scaffold a new ArcKit project with Copilot prompt files

arckit init my-project --ai copilot

The arckit init command, implemented in scripts/python/create-project.py, creates:

• .github/prompts/ directory with 68 .prompt.md files
• .github/agents/ folder with 10 custom agents
• copilot-instructions.md for repository-wide context

Install Arc-Kit for Codex CLI or OpenCode CLI

For Codex CLI or OpenCode CLI, install using pip or the faster uv tool. These options are documented in README lines 68-77【/blob/main/README.md#L68-L77】:

# Standard pip installation

pip install git+https://github.com/tractorjuice/arc-kit.git

# Alternative with uv (faster resolution)

uv tool install arckit-cli --from git+https://github.com/tractorjuice/arc-kit.git

# Or run without installing

uvx --from git+https://github.com/tractorjuice/arc-kit.git arckit init my-project

After installation, run arckit init to scaffold the project structure (.arckit/, .codex/, .agents/skills/, etc.). The generated skills are auto-discovered by Codex CLI using the $arckit-* naming pattern.

Verify Your Installation

Confirm Arc-Kit installed correctly using the CLI commands defined in src/arckit_cli/__init__.py【/blob/main/src/arckit_cli/init.py#L1-L200】:

arckit --help          # Displays CLI help text

arckit check           # Validates required tools (typer, rich, etc.)

The check command verifies that all dependencies listed in pyproject.toml are available in your environment.

Summary

• Arc-Kit installs via marketplace plugins for Claude Code and Gemini CLI, or via pip/uv for GitHub Copilot and Codex CLI.
• Python 3.11+ is required as specified in pyproject.toml.
• arckit init scaffolds platform-specific configurations (prompts, agents, instructions) using logic from scripts/python/create-project.py.
• 68 slash commands and 10 autonomous agents become available immediately after installation.
• Auto-updates are handled by the marketplace for Claude Code and Gemini CLI; pip-based installs require manual updates.

Frequently Asked Questions

How do I update Arc-Kit after installation?

For Claude Code and Gemini CLI, updates are automatic when the marketplace repository changes. For pip-based installations (GitHub Copilot, Codex CLI), run pip install --upgrade git+https://github.com/tractorjuice/arc-kit.git or use uv tool upgrade arckit-cli if you installed with uv.

Can I install Arc-Kit on Python versions older than 3.11?

No. Arc-Kit requires Python 3.11 or higher as explicitly declared in the pyproject.toml requires-python field. Attempting to install on older versions will result in a dependency resolution error from pip or uv.

What is the difference between arckit init and the marketplace plugin?

The marketplace plugin (for Claude Code and Gemini CLI) provides immediate access to Arc-Kit's commands and agents within the AI assistant's native environment without local project scaffolding. arckit init (for Copilot and Codex CLI) creates a local project structure with prompt files, agent definitions, and configuration files that the AI assistant reads from your repository.

Where can I find the source code for the Arc-Kit CLI commands?

The CLI entry point and command implementations are located in src/arckit_cli/__init__.py. The project scaffolding logic used by arckit init resides in scripts/python/create-project.py. Both files are part of the tractorjuice/arc-kit repository and can be examined to understand how commands are registered and how project templates are generated.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}