Configuration Options in Agent OS

Agent OS uses flexible configuration options that allow you to customize how it works with your AI coding tools. Instead of rigid "modes", you can now mix and match features to fit your workflow.

Configuration Options

claude_code_commands (true/false)

Enable this to use Agent OS with Claude Code slash commands. When enabled, you can run Agent OS workflows using commands like /shape-spec, /write-spec, /create-tasks, etc.

Set this to true if you're using Claude Code as your primary AI coding tool.

use_claude_code_subagents (true/false)

Enable this to allow Claude Code commands to delegate tasks to specialized subagents. This provides more efficient, hands-off development but comes with tradeoffs like higher token usage and less transparency.

When false, Claude Code commands still work but execute workflows directly without delegating to subagents.

Note: This option only applies when claude_code_commands is true.

agent_os_commands (true/false)

Enable this to generate Agent OS workflow commands in your agent-os/commands/ folder. These can be used as sequential prompts in any AI coding tool (Cursor, Codex, Gemini, Windsurf, etc.).

Set this to true if you're using AI coding tools other than Claude Code.

standards_as_claude_code_skills (true/false)

Enable this to leverage Claude Code Skills integration for intelligent standards application. When enabled, your standards are converted to Claude Code Skills that Claude automatically applies when relevant, providing greater context efficiency.

Note: This option only applies when claude_code_commands is true. This is enabled by default.

Learn more about Skills integration

Setting Your Configuration

Configure these options in your base installation's ~/agent-os/config.yml file:

Copy

defaults:
  profile: default
  claude_code_commands: true
  use_claude_code_subagents: true
  agent_os_commands: false
  standards_as_claude_code_skills: true

You can also override these settings during project installation using command-line flags.

Common Configurations

Claude Code with subagent delegation

Full-featured Claude Code setup with autonomous subagents handling specialized tasks.

Copy

claude_code_commands: true
use_claude_code_subagents: true
standards_as_claude_code_skills: true
agent_os_commands: false

Claude Code without subagent delegation

Use Claude Code commands but execute workflows directly without delegating to subagents.

Copy

claude_code_commands: true
use_claude_code_subagents: false
standards_as_claude_code_skills: true
agent_os_commands: false

Other AI coding tools (Cursor, Codex, etc.)

Generate prompt commands for use with any AI coding tool via @ references.

Copy

claude_code_commands: false
use_claude_code_subagents: false
standards_as_claude_code_skills: true
agent_os_commands: true

Claude Code with Skills

Leverage Skills integration for automatic, context-efficient standards application. This is the recommended default configuration.

Copy

claude_code_commands: true
use_claude_code_subagents: true
agent_os_commands: false
standards_as_claude_code_skills: true

Dual setup (Claude Code + other tools)

Enable both Claude Code commands and agent-os commands for flexibility.

Copy

claude_code_commands: true
use_claude_code_subagents: true
standards_as_claude_code_skills: true
agent_os_commands: true

Goals of Agent OS Configurations

Flexibility - Mix and match features to fit your exact workflow
Tool agnostic - Works with Claude Code, Cursor, or any AI coding tool
Subagent control - Use Claude Code with or without delegating to subagents
Standards integration - Choose how standards are delivered (Skills or traditional injection)

Next steps

Install Agent OS

Learn about the workflow

Configuration Options