AI Commands

Use these commands when a coding agent needs current Agentuity CLI context before it edits files or runs project commands. They print context and schemas. They do not create project files by themselves.

agentuity ai intro
agentuity ai capabilities show
agentuity ai schema show
agentuity ai prompt llm

Which command to use

Agent Introduction

Print the short primer a coding agent should read before it modifies an Agentuity project:

agentuity ai intro

The output is framework-first. It tells the agent to use normal create, import, dev, build, and deploy commands, and it treats deploy as a cloud mutation that needs approval.

Use it at the start of an agent-driven setup or when a coding agent joins an existing app.

Capabilities

Show the CLI's agent-facing capabilities:

agentuity ai capabilities show
agentuity --json ai capabilities show

Use JSON when another program needs to parse the result. The output lists capability groups and workflows, but the exact command flags still come from command help or schema output.

Schema Inspection

Use the CLI schema when an agent needs exact command names, options, defaults, and requirements:

agentuity ai schema show
agentuity --help=json

Generate the project config schema when an editor or validation tool needs the agentuity.json shape:

agentuity ai schema generate

Prompt Generation

Generate a broad CLI safety prompt for LLM agents:

agentuity ai prompt llm

Agentic Setup Token

The app can launch an authenticated first-run flow by calling the hidden setup command with a one-time token:

agentuity setup --setup-token <token>

When the token is valid, the command authenticates through the CLI and prints the onboarding prompt used by coding agents. The token must come from the app-to-agent handoff flow. Do not invent one in a prompt.

Integration Examples

IDE Extension

Use the CLI to provide autocomplete, validation, and project configuration help:

import { execFileSync } from 'node:child_process';
 
const capabilities: unknown = JSON.parse(
  execFileSync('agentuity', ['--json', 'ai', 'capabilities', 'show'], {
    encoding: 'utf8',
  })
);

import { execFileSync } from 'node:child_process';
 
const schema: unknown = JSON.parse(
  execFileSync('agentuity', ['ai', 'schema', 'generate'], {
    encoding: 'utf8',
  })
);

AI Coding Agent

agentuity ai intro > /tmp/agentuity-context.md
cat /tmp/agentuity-context.md

OpenCode Plugin

Install the Agentuity plugin for OpenCode to enable AI-assisted development:

# Install the plugin
agentuity ai opencode install
 
# Remove the plugin
agentuity ai opencode uninstall
 
# Run a task in headless mode
agentuity ai opencode run "Create an agent that validates email addresses"
 
# Run in a cloud sandbox environment
agentuity ai opencode run --sandbox "run the test suite"
 
# Output raw JSON events (for programmatic consumption)
agentuity ai opencode run --json "fix the bug"

The plugin provides Agentuity Coder agents, slash commands, memory support, and session tools. See OpenCode Plugin for full documentation.

Claude Code Plugin

Install or remove the Agentuity Coder plugin for Claude Code:

agentuity ai claude-code install
agentuity ai claude-code uninstall

The installer adds the plugin package locally, configures Claude Code permissions, and prints the claude --plugin-dir ... command to start Claude Code with the plugin. See Claude Code Plugin for setup details.

Cadence Mode

Cadence enables long-running autonomous work sessions. Access it via:

• Interactive: Use the /agentuity-cadence slash command within OpenCode
• Headless: Run agentuity ai opencode run "/agentuity-cadence <task>" from the CLI

# Start a Cadence session interactively in OpenCode
/agentuity-cadence Build a complete auth system with tests
 
# Or run headlessly from the CLI
agentuity ai opencode run "/agentuity-cadence Build a complete auth system with tests"

See OpenCode Plugin: Cadence Mode for details on controlling Cadence sessions.

Next Steps

• Getting Started with the CLI: Install and authenticate
• Local Development: Run Agentuity projects locally
• Agents: Structure model-backed workflows in current Agentuity apps