AIWG CLI Reference

Complete reference for all aiwg CLI commands.

Prerequisites: Node.js ≥18.0.0 and npm install -g aiwg

References:

@src/extensions/commands/definitions.ts - Command extension definitions
@src/extensions/types.ts - Extension type system
@.aiwg/architecture/unified-extension-schema.md - Extension schema documentation

Maintenance Commands
Framework Management
Project Setup
Workspace Management
MCP Commands
Catalog Commands
Toolsmith Commands
Utility Commands
Plugin Commands
Scaffolding Commands
Mission Control Commands
Agent Team Commands
Ralph Commands
Documentation Commands
SDLC Orchestration Commands
Index Commands
Addon Commands

Maintenance Commands

help

Display comprehensive CLI help information.

aiwg help
aiwg -help
aiwg --help

Capabilities: cli, help, documentation Platforms: All Tools: None required

Shows:

Available commands grouped by category
Common usage patterns
Platform-specific notes
Links to documentation

version

Show version and channel information.

aiwg version
aiwg -version
aiwg --version

Capabilities: cli, version, info Platforms: All Tools: Read

Shows:

Current AIWG version
Active channel (stable/main)
Installation path
Node.js version

Example output:

AIWG v2026.1.5 (stable)
Installed: ~/.nvm/versions/node/v20.10.0/lib/node_modules/aiwg
Node.js: v20.10.0

doctor

Check installation health and diagnose issues.

aiwg doctor

Capabilities: cli, diagnostics, health-check Platforms: All Tools: Read, Bash

Checks:

AIWG installation and version
Node.js version compatibility
Project .aiwg/ directory structure
Framework registry status
Deployed agents and commands
MCP server availability
System dependencies (git, jq, etc.)

Example output:

✓ AIWG installed: v2026.1.5
✓ Node.js version: v20.10.0 (meets requirement ≥18.0.0)
✓ Project directory: /home/user/my-project
✓ Framework registry: 2 frameworks installed
✓ Agents deployed: 15
✓ Commands deployed: 31
⚠ MCP server not configured
ℹ Run 'aiwg mcp install claude' to configure MCP

update

Check for and apply updates.

aiwg update
aiwg -update

Capabilities: cli, update, maintenance Platforms: All Tools: Bash

Actions:

Checks npm registry for newer version
Shows changelog highlights
Prompts for update confirmation
Runs npm update -g aiwg
Verifies successful update

Channel switching:

# Switch to bleeding edge (main branch)
aiwg --use-main

# Switch back to stable
aiwg --use-stable

sync

Sync AIWG to latest version and re-deploy all frameworks to active provider.

aiwg sync
aiwg --sync

Capabilities: cli, sync, maintenance, deploy, self-maintenance Platforms: All Tools: Bash, Read

Actions:

Detects active provider (claude-code, copilot, cursor, etc.)
Checks current AIWG version
Updates package to latest (unless --skip-update)
Re-deploys all installed frameworks (or specific ones)
Runs health check via aiwg doctor

Flags:

Flag	Description
`--dry-run`	Show what would change without making changes
`--quiet`	Machine-readable JSON output (for orchestration)
`--skip-update`	Skip npm update, only re-deploy frameworks
`--provider <name>`	Target specific provider (default: auto-detect)
`--channel <name>`	Update channel (stable, main)
`--frameworks <list>`	Comma-separated frameworks to re-deploy

Examples:

# Full sync (update + re-deploy + verify)
aiwg sync

# Check what would change
aiwg sync --dry-run

# Sync to specific provider
aiwg sync --provider copilot

# Re-deploy only SDLC framework, skip update
aiwg sync --skip-update --frameworks sdlc

# Quiet mode for agent orchestration
aiwg sync --quiet

Example output:

◆ aiwg sync
──────────────────────────────
ℹ Detecting provider...
✓ Provider: claude
ℹ Checking version...
✓ Version check complete
ℹ Checking for updates...
✓ Package up to date
ℹ Re-deploying frameworks...
✓ Deployed: all
ℹ Running health check...
✓ Health check passed
──────────────────────────────
✓ Sync complete

Framework Management

use

Install and deploy framework or addon to your project. Skills are deployed natively for providers that support them; commands are generated from skill sources for providers that need them. Each provider receives all artifact types (agents, skills, commands, rules) regardless of native platform support.

aiwg use <framework|addon>

Arguments:

<framework> - Framework name: sdlc, marketing, writing, all
<addon> - Addon name: rlm

Options:

--provider <name> - Target platform (claude, copilot, factory, cursor, windsurf, warp, codex, opencode, hermes, openclaw, local)
--model <name> - Override model for all tiers (blanket)
--reasoning-model <name> - Override reasoning tier model (alias: --reasoning)
--coding-model <name> - Override coding tier model (alias: --coding)
--efficiency-model <name> - Override efficiency tier model (alias: --efficiency)
--save - Save model overrides to project models.json
--save-user - Save model overrides to ~/.config/aiwg/models.json
--no-utils - Skip aiwg-utils addon installation (frameworks only)
--force - Overwrite existing deployments
--dry-run - Preview without making changes

Capabilities: cli, framework, deployment, addon Platforms: All Tools: Read, Write, Bash, Glob

Examples:

# Deploy SDLC framework for Claude Code (default)
aiwg use sdlc

# Deploy to GitHub Copilot
aiwg use sdlc --provider copilot

# Deploy marketing framework
aiwg use marketing

# Deploy all frameworks
aiwg use all

# Deploy RLM addon (recursive context decomposition)
aiwg use rlm

# Deploy RLM addon to Codex
aiwg use rlm --provider codex

# Preview deployment without writing files
aiwg use sdlc --dry-run

# Override model for all tiers
aiwg use sdlc --model sonnet

# Override individual tiers
aiwg use sdlc --reasoning opus --coding sonnet --efficiency haiku

# Use a specific model ID on Factory
aiwg use sdlc --provider factory --coding-model gpt-5.3-codex

# Blanket with per-tier override
aiwg use sdlc --model sonnet --reasoning opus

# Save model overrides for future deployments
aiwg use sdlc --model sonnet --save

Model override precedence: CLI flags > project models.json > user ~/.config/aiwg/models.json > AIWG defaults

Shorthand values: opus, sonnet, haiku, inherit — resolved per provider to full model IDs

Framework options:

Framework	ID	Description
SDLC Complete	`sdlc`	Full software development lifecycle with 90 agents
Marketing Kit	`marketing`	Complete marketing campaign management
Writing Quality	`writing`	Voice profiles and content validation
All	`all`	Deploy all frameworks

Addon options:

Addon	ID	Description
RLM	`rlm`	Recursive Language Models — recursive context decomposition for 10M+ token processing

Platform targets:

Platform	`--provider` ID	Artifact dirs	Behaviors
Claude Code	`claude`	`.claude/agents/`, `.claude/commands/`, `.claude/skills/`, `.claude/rules/`	—
GitHub Copilot	`copilot`	`.github/agents/`, `.github/copilot-rules/`, `.github/skills/`	—
Factory AI	`factory`	`.factory/droids/`, `.factory/commands/`, `.factory/skills/`, `.factory/rules/`	—
Cursor	`cursor`	`.cursor/agents/`, `.cursor/commands/`, `.cursor/skills/`, `.cursor/rules/`	—
Windsurf	`windsurf`	`AGENTS.md` (aggregated), `.windsurf/workflows/`, `.windsurf/skills/`, `.windsurf/rules/`	—
Warp Terminal	`warp`	`.warp/agents/`, `.warp/commands/`, `.warp/skills/`, `.warp/rules/`, `WARP.md` (aggregated)	—
OpenAI/Codex	`codex`	`.codex/agents/`, `~/.codex/prompts/`, `~/.codex/skills/`, `.codex/rules/`	—
OpenCode	`opencode`	`.opencode/agent/`, `.opencode/commands/`, `.opencode/skill/`, `.opencode/rule/`	—
Hermes	`hermes`	`~/.hermes/skills/`, `AGENTS.md` (lean)	—
OpenClaw	`openclaw`	`~/.openclaw/agents/`, `~/.openclaw/commands/`, `~/.openclaw/skills/`, `~/.openclaw/rules/`, `~/.openclaw/behaviors/`	✓
Local/Ollama	`local`	Same as `claude` (local model, Claude Code paths)	—

Notes:

Codex: Commands and skills deploy to ~ (user-level) for availability across all projects; the provider ID is codex, not openai
Windsurf: Agents aggregated into AGENTS.md at project root; no separate agent files
Warp: Agents and commands also aggregated into WARP.md for single-file context loading
Hermes: Not a spawnable CLI — access via ollama run hermes3 or MCP sidecar; deploy sets up skills and a lean AGENTS.md
OpenClaw: Only provider with behaviors support (~/.openclaw/behaviors/); all artifacts deploy to home directory
Local/Ollama: Uses Claude Code path layout; specify --coding-model ollama/<model> to route coding tasks to the local model

list

List installed frameworks and addons.

aiwg list

Capabilities: cli, framework, query Platforms: All Tools: Read

Output format:

Installed Frameworks:
  sdlc-complete (v1.0.0) - Full SDLC framework
  media-marketing-kit (v1.0.0) - Marketing framework

Installed Addons:
  aiwg-utils (v1.0.0) - Core utilities
  voice-framework (v1.0.0) - Voice profiles

Total: 2 frameworks, 2 addons

remove

Remove a framework or addon.

aiwg remove <id>

Arguments:

<id> - Framework or addon ID (e.g., sdlc, marketing, voice-framework)

Capabilities: cli, framework, uninstall Platforms: All Tools: Read, Write, Bash

Examples:

# Remove SDLC framework
aiwg remove sdlc

# Remove voice framework addon
aiwg remove voice-framework

Actions:

Removes deployed files from .claude/, .github/, etc.
Updates framework registry
Removes workspace artifacts (.aiwg/<framework>/)
Preserves user-created content

Project Setup

new

Create new project with SDLC templates.

aiwg new <project-name>
aiwg -new <project-name>

Arguments:

<project-name> - Name of project directory to create

Capabilities: cli, project, scaffolding Platforms: All Tools: Read, Write, Bash

Creates:

my-project/
├── .aiwg/
│   ├── intake/
│   ├── requirements/
│   ├── architecture/
│   ├── planning/
│   ├── risks/
│   ├── testing/
│   ├── security/
│   ├── deployment/
│   └── frameworks/
│       └── registry.json
├── .claude/
│   ├── agents/
│   ├── commands/
│   └── skills/
├── CLAUDE.md
└── README.md

Example:

aiwg new customer-portal
cd customer-portal

# Framework already deployed, start working
/intake-wizard "Customer portal with real-time chat"

Workspace Management

status

Show workspace health and installed frameworks.

aiwg status
aiwg -status

Capabilities: cli, workspace, status Platforms: All Tools: Read, Bash

Shows:

Project directory
Installed frameworks and versions
Framework health status
Agent deployment count
Command deployment count
Workspace artifact summary
Git status (if git repo)

Example output:

Workspace: /home/user/customer-portal
Git: clean (main branch)

Frameworks:
  ✓ sdlc-complete v1.0.0 (90 agents, 42 commands)
  ✓ aiwg-utils v1.0.0

Artifacts:
  Requirements: 12 files
  Architecture: 5 files
  Tests: 8 files

Status: Healthy

migrate-workspace

Migrate legacy .aiwg/ to framework-scoped structure.

aiwg migrate-workspace

Capabilities: cli, workspace, migration Platforms: All Tools: Read, Write, Bash

Migrates:

From (legacy):

.aiwg/
├── intake/
├── requirements/
└── ...

To (framework-scoped):

.aiwg/
├── frameworks/
│   ├── registry.json
│   └── sdlc-complete/
│       ├── intake/
│       ├── requirements/
│       └── ...
└── shared/

Safety:

Creates backup in .aiwg.backup-<timestamp>/
Validates migration before committing
Preserves all content
Updates framework registry

rollback-workspace

Rollback workspace migration from backup.

aiwg rollback-workspace

Capabilities: cli, workspace, rollback Platforms: All Tools: Read, Write, Bash

Restores from:

.aiwg.backup-<timestamp>/ directories
Prompts to select backup if multiple exist
Validates backup before restoring
Creates pre-rollback backup

MCP Commands

mcp

MCP server operations.

aiwg mcp <subcommand>

Subcommands:

mcp serve

Start the AIWG MCP server.

aiwg mcp serve

Actions:

Starts stdio-based MCP server
Exposes AIWG tools, resources, and prompts
Supports Claude Desktop, Cursor, Factory

mcp install

Generate MCP client configuration.

aiwg mcp install <client>

Arguments:

<client> - Client name: claude, cursor, factory

Options:

--dry-run - Preview without writing

Actions:

Generates client-specific config
Adds to ~/.config/claude/config.json (Claude Desktop)
Adds to .cursor/config.json (Cursor)
Shows manual steps if auto-install fails

Example:

# Install for Claude Desktop
aiwg mcp install claude

# Preview config
aiwg mcp install claude --dry-run

mcp info

Show MCP server capabilities.

aiwg mcp info

Shows:

MCP protocol version
Available tools
Available resources
Available prompts
Server status

Capabilities: cli, mcp, server Platforms: All Tools: Read, Write, Bash

Catalog Commands

catalog

Model catalog operations.

aiwg catalog <subcommand>

Subcommands:

catalog list

List available models.

aiwg catalog list

Options:

--provider <name> - Filter by provider (anthropic, openai, google)
--type <type> - Filter by type (chat, completion, embedding)

catalog info

Show model information.

aiwg catalog info <model-id>

Arguments:

<model-id> - Model identifier (e.g., claude-opus-4-5-20251101)

catalog search

Search model catalog.

aiwg catalog search <query>

Arguments:

<query> - Search terms

Capabilities: cli, catalog, models Platforms: All Tools: Read

Toolsmith Commands

runtime-info

Show runtime environment summary with tool discovery.

aiwg runtime-info

Capabilities: cli, toolsmith, discovery Platforms: All Tools: Read, Bash

Shows:

Platform detection (Claude Code, Cursor, etc.)
Available tools (Read, Write, Bash, Glob, Grep)
System utilities (git, jq, curl, etc.)
Environment variables
Tool capabilities and limitations

Example output:

Platform: Claude Code
AI Model: claude-sonnet-4-5-20250929

Available Tools:
  ✓ Read (supports images, PDFs)
  ✓ Write
  ✓ Bash (timeout: 2min)
  ✓ Glob
  ✓ Grep

System Utilities:
  ✓ git v2.39.0
  ✓ jq v1.6
  ✓ node v20.10.0
  ✓ npm v10.2.3
  ✗ gh (GitHub CLI not installed)

Scheduler:
  Backend:  native-cron (CronCreate) / aiwg-cli fallback
  Chrony:   ✓ installed (precise NTP)

Environment: Linux 6.14.0-37-generic

Schedule Skill

Cross-provider scheduler that detects native cron capability (Claude Code CronCreate) and falls back to the AIWG daemon CLI on all other providers. Checks chrony installation for precise timing.

schedule create

/schedule create --name <name> --cron "<expr>" --task "<prompt>"

Options:

--name — Unique task name (required)
--cron — 5-field cron expression (required)
--task — Prompt or command to run (required)
--provider native|aiwg-cli — Override backend detection

Examples:

/schedule create --name daily-sync --cron "0 9 * * *" --task "aiwg sync"
/schedule create --name health-check --cron "0 */6 * * *" --task "aiwg doctor"

schedule list

/schedule list

Lists all scheduled tasks, showing name, cron expression, next run time, and backend in use.

schedule delete

/schedule delete --name <name>

Deletes a scheduled task by name.

Backend Detection

Provider	Backend
Claude Code	`native-cron` (CronCreate/CronList/CronDelete)
All others	`aiwg-cli` (AIWG daemon)

The active backend is reported in aiwg runtime-info. Override with --provider flag.

Chrony Recommendation

When scheduling tasks, the skill checks whether chrony is installed and recommends it if missing. Chrony provides accurate NTP time synchronization, preventing clock drift that causes tasks to run at unexpected times — especially on servers that sleep or in virtual environments.

# Ubuntu/Debian
sudo apt install chrony

# RHEL/Fedora
sudo dnf install chrony

# macOS
brew install chrony

Utility Commands

prefill-cards

Prefill SDLC card metadata from team profile.

aiwg prefill-cards

Capabilities: cli, sdlc, automation Platforms: All Tools: Read, Write

Actions:

Reads .aiwg/team-profile.json
Finds empty SDLC cards (use cases, architecture docs, etc.)
Fills in standard metadata (author, date, version)
Preserves existing content

Example:

# Create team profile first
cat > .aiwg/team-profile.json <<EOF
{
  "project": "Customer Portal",
  "team": "Platform Team",
  "defaultAuthor": "Jane Developer",
  "defaultReviewer": "John Architect"
}
EOF

# Prefill all cards
aiwg prefill-cards

contribute-start

Start AIWG contribution workflow.

aiwg contribute-start

Capabilities: cli, contribution, workflow Platforms: All Tools: Read, Write, Bash

Actions:

Guides through contribution setup
Creates feature branch
Sets up development environment
Links to contribution guidelines

validate-metadata

Validate plugin/agent metadata.

aiwg validate-metadata [path]

Arguments:

[path] - Optional path to validate (defaults to current directory)

Capabilities: cli, validation, metadata Platforms: All Tools: Read

Validates:

Extension schema compliance
Required fields present
Version format correct
Platform compatibility declared
Keywords and capabilities present

Example:

# Validate all extensions in current directory
aiwg validate-metadata

# Validate specific extension
aiwg validate-metadata .claude/agents/api-designer.md

Plugin Commands

Note: Plugin commands are specific to Claude Code integration.

install-plugin

Install Claude Code plugin.

aiwg install-plugin <name>

Arguments:

<name> - Plugin name from marketplace

Capabilities: cli, plugin, install Platform: Claude Code only Tools: Read, Write, Bash

Example:

aiwg install-plugin sdlc@aiwg

uninstall-plugin

Uninstall Claude Code plugin.

aiwg uninstall-plugin <name>

Arguments:

<name> - Plugin name

Capabilities: cli, plugin, uninstall Platform: Claude Code only Tools: Read, Write, Bash

plugin-status

Show Claude Code plugin status.

aiwg plugin-status

Capabilities: cli, plugin, status Platform: Claude Code only Tools: Read

Shows:

Installed plugins
Plugin versions
Enabled/disabled status
Marketplace connection

package-plugin

Package specific plugin for Claude Code marketplace.

aiwg package-plugin <name>

Arguments:

<name> - Plugin name to package

Capabilities: cli, plugin, packaging Platforms: Claude Code, Generic Tools: Read, Write, Bash

Creates:

dist/plugins/<name>.plugin.tar.gz
Manifest validation
README and LICENSE inclusion

package-all-plugins

Package all plugins for Claude Code marketplace.

aiwg package-all-plugins

Capabilities: cli, plugin, packaging Platforms: Claude Code, Generic Tools: Read, Write, Bash

Creates:

Packages for: sdlc, marketing, utils, voice
Validates all manifests
Generates marketplace index

Scaffolding Commands

Commands for creating new extensions within addons/frameworks.

Skills vs Commands — Provider Support

Skills are the canonical source type for agentic workflows. During aiwg use deployment:

Provider support	Behavior
Native skill support (Claude Code, OpenCode, Warp, etc.)	Skill deployed as-is to `.{platform}/skills/{id}/SKILL.md`
Generated-command providers (Copilot, Factory, etc.)	Command file generated from skill source, deployed alongside skill
Legacy direct commands	Authored command files still supported; not generated from a skill

Authoring guidance:

New workflow? → aiwg add-skill — AIWG handles deployment and command generation
Modifying an existing workflow? → Edit the SKILL.md source, not the generated command files
Advanced direct command? → aiwg add-command (deprecated path, still supported)

add-agent

Add agent to addon/framework.

aiwg add-agent <name>

Arguments:

<name> - Agent name (e.g., "API Designer")

Capabilities: cli, scaffolding, agent Platforms: All Tools: Read, Write

Creates:

Agent markdown file with frontmatter
Extension definition entry
Platform-specific adaptations

Example:

aiwg add-agent "API Designer"

Creates: agents/api-designer.md

add-command

Deprecated: Use aiwg add-skill instead. Skills are the primary workflow extension type; commands are generated from skills during deployment. add-command remains available for direct command authoring in advanced cases.

Add command to addon/framework.

aiwg add-command <name>

Arguments:

<name> - Command name (e.g., "validate-api")

Capabilities: cli, scaffolding, command Platforms: All Tools: Read, Write

add-skill

Add skill to addon/framework.

aiwg add-skill <name>

Arguments:

<name> - Skill name (e.g., "project-awareness")

Capabilities: cli, scaffolding, skill Platforms: All Tools: Read, Write

add-behavior

Scaffold a new behavior with BEHAVIOR.md and scripts.

aiwg add-behavior <name> [options]

Arguments:

<name> - Behavior name (kebab-case recommended)

Options:

--description, -d - Behavior description
--hooks - Comma-separated hook events (default: on_file_write). Available: on_file_write, on_tool_complete, on_schedule, on_commit, on_pr_open, on_deploy, on_session_start, on_session_end
--category - Behavior category (default: general)
--dry-run, -n - Preview what would be created

Capabilities: cli, scaffolding, behavior Platforms: Claude Code, OpenClaw Tools: Read, Write

Creates:

agentic/code/behaviors/<name>/
├── BEHAVIOR.md          # Pre-filled with hooks and triggers
└── scripts/
    └── main.sh          # Entry point stub

Examples:

aiwg add-behavior security-scanner
aiwg add-behavior test-watcher --hooks on_file_write,on_schedule --category testing
aiwg add-behavior deploy-guard --hooks on_deploy --description "Pre-deploy validation"

add-template

Add template to addon/framework.

aiwg add-template <name>

Arguments:

<name> - Template name (e.g., "use-case-template")

Capabilities: cli, scaffolding, template Platforms: All Tools: Read, Write

scaffold-addon

Create new addon package.

aiwg scaffold-addon <name>

Arguments:

<name> - Addon name (e.g., "my-addon")

Capabilities: cli, scaffolding, addon Platforms: All Tools: Read, Write

Creates:

agentic/code/addons/my-addon/
├── manifest.json
├── README.md
├── agents/
├── commands/
├── skills/
└── templates/

scaffold-extension

Create new extension package.

aiwg scaffold-extension <name>

Arguments:

<name> - Extension name

Capabilities: cli, scaffolding, extension Platforms: All Tools: Read, Write

scaffold-framework

Create new framework package.

aiwg scaffold-framework <name>

Arguments:

<name> - Framework name (e.g., "security-framework")

Capabilities: cli, scaffolding, framework Platforms: All Tools: Read, Write

Creates:

agentic/code/frameworks/security-framework/
├── manifest.json
├── README.md
├── agents/
├── commands/
├── skills/
├── templates/
└── docs/

Daemon Commands

Commands for managing the AIWG daemon and its subsystems.

behavior

Manage behavior YAML bundles that bind directives and toolsets to agent types.

aiwg behavior <list|info|apply|remove> [name] [options]

Subcommands:

list - List all available behaviors
info <name> - Show behavior details (BEHAVIOR.md content)
apply <name> - Apply a behavior to the daemon
remove <name> - Remove a behavior from the daemon

Capabilities: cli, behavior, daemon, configuration Platforms: Claude Code Tools: Read, Bash, Write

Examples:

aiwg behavior list
aiwg behavior info security-sentinel

daemon-init

Initialize daemon config from a profile template.

aiwg daemon-init [profile-name] [--force]

Arguments:

[profile-name] - Profile template to use (default: manager)

Options:

--force - Overwrite existing config

Capabilities: cli, daemon, configuration, scaffolding Platforms: Claude Code Tools: Bash, Read, Write

Creates:

.aiwg/daemon.yaml from the selected profile template
.env.example with required environment variables

Mission Control Commands

Mission Control provides multi-loop background orchestration for parallel long-running agents.

mc

Multi-loop background orchestration dashboard.

aiwg mc <subcommand> [options]
aiwg mission-control <subcommand> [options]

Capabilities: cli, orchestration, ralph, background, multi-loop, mission-control Platforms: All Tools: Bash, Read, Write

Subcommands:

Subcommand	Description
`start`	Start a new Mission Control session
`dispatch <id> "<objective>"`	Add a background mission to session
`status [<id>] [--json]`	View mission status dashboard
`watch [<id>]`	Live monitor (streaming)
`abort <session> <mission>`	Abort a specific mission
`pause [<id>]`	Pause active session
`resume [<id>]`	Resume paused session
`stop [<id>] [--drain]`	Shut down session
`list [--json]`	List all sessions

Examples:

# Start a named session
aiwg mc start --name "Construction Sprint 4"

# Dispatch missions
aiwg mc dispatch mc-abc123 "Fix auth service" --completion "tests pass" --priority high
aiwg mc dispatch mc-abc123 "Add pagination" --completion "paginated responses"

# Monitor
aiwg mc status mc-abc123
aiwg mc status mc-abc123 --json

# Drain and stop (let running missions finish)
aiwg mc stop mc-abc123 --drain

Example output:

◆ MISSION CONTROL — Construction Sprint 4  [mc-abc123]
──────────────────────────────────────────────────────────
  #    Mission                       Status       Loop     Started
──────────────────────────────────────────────────────────
  1    Fix auth service              ✓ DONE       4/10     14:22
  2    Add pagination                ⏳ RUNNING   3/10     14:25
  3    Write integration tests       ⏺ QUEUED     —        —
──────────────────────────────────────────────────────────
  3 missions  |  1 done  |  1 running  |  1 queued  |  0 failed

State persistence: Session state is stored in .aiwg/ralph-external/mc/sessions/ and survives context resets.

Agent Team Commands

Agent teams provide a provider-agnostic abstraction for multi-agent collaboration. On Claude Code, teams use native agent dispatch. On all other providers (Copilot, Cursor, Warp, Windsurf, OpenCode, Factory, Codex, OpenClaw), teams are emulated via aiwg mc (Mission Control) orchestration.

team

Multi-agent team orchestration across all providers.

aiwg team <subcommand> [options]
aiwg teams <subcommand> [options]

Capabilities: orchestration, agent-teams, multi-provider, mission-control Platforms: All (native on Claude Code, emulated via aiwg mc on others) Category: orchestration

Subcommands

Subcommand	Description
`run <name>`	Execute a team workflow
`list`	List available teams
`info <name>`	Show team definition and roster

Provider Routing

Provider	Backend	Behavior
Claude Code	Native	@agent-name dispatch instructions
Warp, Copilot, Cursor, Windsurf, OpenCode, Factory, Codex, OpenClaw	`aiwg mc` emulation	Generates `mc start` + `mc dispatch` commands

Options

Option	Description
`--provider <p>`	Override provider detection
`--objective "<text>"`	Objective string passed to mc dispatch agents
`--json`	Machine-readable output

Examples

# Run a team (auto-detects provider)
aiwg team run sdlc-review

# Run with explicit provider override
aiwg team run sdlc-review --provider cursor

# Run with custom objective
aiwg team run security-review --objective "Pre-release audit for SOC2"

# List all available teams
aiwg team list

# Machine-readable team list
aiwg team list --json

# Inspect team definition
aiwg team info sdlc-review
aiwg team info api-development --json

Built-in Teams (sdlc-complete framework)

Team	Agents	Dispatch	Best For
`api-development`	4	sequential	API design and implementation
`full-stack`	4	sequential	Full-stack feature delivery
`greenfield`	4	sequential	New project setup
`maintenance`	4	sequential	Code review and bug fixing
`migration`	4	sequential	Technology migrations
`sdlc-review`	4	parallel	Phase gate validation
`security-review`	3	sequential	Security audits

Team Definition Format

Teams are defined as JSON files (with an optional dispatch field for parallel | sequential | consensus):

{
  "name": "SDLC Review Team",
  "slug": "sdlc-review",
  "description": "Full SDLC phase gate review team",
  "dispatch": "parallel",
  "agents": [
    { "agent": "security-architect", "role": "reviewer" },
    { "agent": "test-architect",     "role": "reviewer" },
    { "agent": "requirements-analyst", "role": "reviewer" },
    { "agent": "technical-writer",   "role": "reviewer" }
  ],
  "use_cases": ["Phase gate validation", "Release readiness review"],
  "sdlc_phases": ["inception", "elaboration", "construction", "transition"]
}

Custom teams can be placed in .aiwg/teams/<slug>.json for project-local overrides.

Source: agentic/code/frameworks/sdlc-complete/teams/ Schema: agentic/code/frameworks/sdlc-complete/teams/schema.json

Ralph Commands

Ralph is the iterative task execution loop with advanced control layers (Epic #26).

ralph

Start Ralph task execution loop.

aiwg ralph "<task-description>"

Arguments:

<task-description> - Natural language task description

Options:

Core Options:

--completion "<criteria>" - Success criteria (e.g., "npm test passes")
--max-iterations <n> - Maximum iterations (default: 10)
--timeout <seconds> - Per-iteration timeout (default: 300)
--provider <name> - CLI provider: claude (default), codex, opencode, local
--budget <usd> - Budget per iteration in USD (default: 2.0)
--gitea-issue - Create/link Gitea issue for tracking
--mcp-config <json> - MCP server configuration JSON

Research-Backed Options (REF-015, REF-021):

-m, --memory <n|preset> - Memory capacity Ω: 1-10 or preset (simple, moderate, complex, maximum). Default: 3
--cross-task / --no-cross-task - Enable/disable cross-task learning (default: enabled)
--no-analytics - Disable iteration analytics
--no-best-output - Disable best output selection (use final iteration)
--no-early-stopping - Disable early stopping on high confidence

Epic #26 Control Options:

--enable-pid-control - Enable PID control layer (default: true)
--disable-pid-control - Disable PID control layer
--enable-overseer - Enable oversight layer (default: true)
--disable-overseer - Disable oversight layer
--enable-semantic-memory - Enable cross-loop memory (default: true)
--disable-semantic-memory - Disable cross-loop memory
--gain-profile <name> - PID gain profile: conservative, standard, aggressive, recovery, cautious (default: standard)
--validation-level <level> - Validation strictness: minimal, standard, strict (default: standard)
--intervention-mode <mode> - Oversight intervention mode: permissive, balanced, strict (default: balanced)

Capabilities: cli, ralph, orchestration Platforms: All Tools: Read, Write, Bash

Examples:

# Basic task execution
aiwg ralph "Fix all failing tests" --completion "npm test passes"

# Conservative run for security fix (Epic #26)
aiwg ralph "Fix SQL injection" \
  --completion "security scan passes" \
  --gain-profile conservative \
  --validation-level strict

# Fast documentation generation (Epic #26)
aiwg ralph "Generate API docs" \
  --completion "docs/ updated" \
  --gain-profile aggressive \
  --disable-overseer

# Leverage cross-loop memory (Epic #26)
aiwg ralph "Fix auth tests" \
  --completion "tests pass" \
  --enable-semantic-memory

# Refactoring with balanced controls
aiwg ralph "Extract common utilities to shared module" \
  --completion "No lint errors" \
  --gain-profile standard \
  --intervention-mode balanced

# Multi-provider: run with Codex
aiwg ralph "Migrate utils to TypeScript" \
  --completion "npx tsc --noEmit exits 0" \
  --provider codex \
  --budget 3.0

# Research-backed: enhanced memory with cross-task learning
aiwg ralph "Fix all integration tests" \
  --completion "npm test passes" \
  --memory complex \
  --cross-task

Iteration pattern:

Analyze current state (with PID control input)
Plan next step (informed by semantic memory)
Execute step
Verify progress (oversight validation)
Check completion criteria
Repeat or finish

Control Layers (Epic #26):

PID Control Layer:

Adjusts agent autonomy based on progress
Prevents oscillation and runaway behavior
Gain profiles optimize for different scenarios:
- conservative: Slow, cautious (Kp=0.3, Ki=0.05, Kd=0.1)
- standard: Balanced (Kp=0.5, Ki=0.1, Kd=0.2) - default
- aggressive: Fast, high autonomy (Kp=0.8, Ki=0.2, Kd=0.3)
- recovery: Designed for error recovery (Kp=0.4, Ki=0.15, Kd=0.25)
- cautious: Extra validation (Kp=0.2, Ki=0.03, Kd=0.05)

Semantic Memory:

Remembers learnings across loop runs
Queries similar past situations
Prevents repeating mistakes
Shares insights between tasks

Oversight Layer:

Validates actions before execution
Flags risky operations
Requires confirmation for critical changes
Intervention modes:
- permissive: Minimal intervention, trust agent
- balanced: Standard safety checks - default
- strict: Maximum oversight, confirm everything

Crash recovery: State saved in .aiwg/ralph/current-loop.json

ralph-status

Show Ralph loop status.

aiwg ralph-status

Capabilities: cli, ralph, status Platforms: All Tools: Read

Shows:

Current loop active/inactive
Task description
Iterations completed
Success criteria
Last state
Completion percentage estimate
Epic #26 status:
- PID control state (current gains, control signal, error metrics)
- Memory layer stats (entries retrieved, last query, similarity scores)
- Oversight status (active interventions, warnings issued, health score)

Example output:

Ralph Loop Status: Active

Task: Fix all failing tests
Iterations: 3/10
Success Criteria: npm test passes

Last Action: Fixed auth service test
State: In progress
Progress: ~40%

=== Epic #26 Control Layers ===

PID Control:
  Gain Profile: standard
  Current Gains: Kp=0.5, Ki=0.1, Kd=0.2
  Control Signal: 0.42 (moderate autonomy)
  Error: -0.15 (slightly below target progress)
  Integral: 0.08
  Derivative: -0.03

Semantic Memory:
  Total Entries: 127
  Last Retrieval: 2 similar situations found
  Top Match: "auth-test-fix-2024-01" (similarity: 0.87)
  Applied Learnings: 3

Oversight:
  Intervention Mode: balanced
  Active Interventions: 1 (validation flag on file deletion)
  Warnings Issued: 0
  Health Score: 0.92 (healthy)

Next: Resume with '/ralph-resume'

ralph-abort

Abort running Ralph loop.

aiwg ralph-abort

Capabilities: cli, ralph, control Platforms: All Tools: Read, Write

Actions:

Stops current loop
Saves final state (including Epic #26 control state)
Archives loop history
Cleans up temporary files
Preserves semantic memory learnings

ralph-resume

Resume paused Ralph loop.

aiwg ralph-resume

Capabilities: cli, ralph, control Platforms: All Tools: Read, Write

Actions:

Loads last saved state (including Epic #26 control layers)
Restores PID controller state
Reloads semantic memory context
Continues from last iteration
Applies same completion criteria
Respects remaining iteration budget

ralph-attach

Attach to a running Ralph loop's live output stream.

aiwg ralph-attach

Capabilities: cli, ralph, control, monitoring Platforms: All Tools: Read

Actions:

Attaches to a running external Ralph loop
Streams live output (press Ctrl+C to detach)
Shows current iteration progress in real time
Does not affect the running loop

ralph-external

Start external Ralph loop with full crash recovery.

aiwg ralph-external "<task-description>"

Arguments:

<task-description> - Natural language task description

Options:

All options from ralph command plus:

External-Specific Options:

--checkpoint-interval <n> - Checkpoint every N iterations (default: 1)
--crash-recovery - Enable crash recovery (default: true)
--state-file <path> - Custom state file location (default: .aiwg/ralph-external/state.json)

Epic #26 Control Options:

Same as ralph command

Capabilities: cli, ralph, orchestration, external Platforms: All Tools: Read, Write, Bash

Examples:

# External loop with crash recovery
aiwg ralph-external "Refactor payment module" \
  --completion "tests pass" \
  --checkpoint-interval 2

# Critical task with strict controls
aiwg ralph-external "Migrate database schema" \
  --completion "migration complete" \
  --gain-profile conservative \
  --validation-level strict \
  --intervention-mode strict \
  --checkpoint-interval 1

Difference from ralph:

Designed for longer-running tasks
Full state persistence to disk
Automatic checkpoint creation
Recoverable across process restarts
Ideal for CI/CD integration

ralph-memory

Manage semantic memory (Epic #26).

aiwg ralph-memory <subcommand>

Subcommands:

ralph-memory list

List all semantic memory learnings.

aiwg ralph-memory list

Options:

--limit <n> - Limit results (default: 20)
--sort <field> - Sort by: date, similarity, usage_count (default: date)

Example output:

Semantic Memory Learnings (127 total)

1. auth-test-fix-2024-01 (2024-01-15)
   Situation: Fixing authentication test failures
   Learning: Check token expiration config first
   Used: 5 times

2. sql-injection-fix-2024-02 (2024-01-20)
   Situation: SQL injection vulnerability
   Learning: Use parameterized queries, not string concat
   Used: 3 times

...

ralph-memory query

Query semantic memory for similar situations.

aiwg ralph-memory query "<pattern>"

Arguments:

<pattern> - Query text or pattern

Options:

--threshold <n> - Similarity threshold 0-1 (default: 0.7)
--limit <n> - Max results (default: 10)

Example:

aiwg ralph-memory query "authentication failing"

ralph-memory prune

Clean old or unused memory entries.

aiwg ralph-memory prune [--older-than <days>]

Options:

--older-than <days> - Remove entries older than N days (default: 90)
--unused - Remove entries never referenced
--dry-run - Preview without deleting

ralph-memory export

Export memory to JSON.

aiwg ralph-memory export <file>

Arguments:

<file> - Output file path

Example:

aiwg ralph-memory export memory-backup.json

ralph-memory import

Import memory from JSON.

aiwg ralph-memory import <file>

Arguments:

<file> - Input file path

Options:

--merge - Merge with existing (default: replace)

Capabilities: cli, ralph, memory Platforms: All Tools: Read, Write

ralph-config

View and configure Epic #26 control layers.

aiwg ralph-config <subcommand>

Subcommands:

ralph-config show

Show current Ralph configuration.

aiwg ralph-config show

Example output:

Ralph Configuration

PID Control:
  Enabled: true
  Gain Profile: standard
  Gains: Kp=0.5, Ki=0.1, Kd=0.2

Semantic Memory:
  Enabled: true
  Database: .aiwg/ralph/memory.db
  Entry Count: 127

Oversight:
  Enabled: true
  Intervention Mode: balanced
  Validation Level: standard

Checkpoints:
  Enabled: true
  Interval: 1 iteration
  Location: .aiwg/ralph/

ralph-config set

Set configuration value.

aiwg ralph-config set <key> <value>

Arguments:

<key> - Configuration key (dot-notation)
<value> - New value

Examples:

# Change gain profile
aiwg ralph-config set pid.gain_profile aggressive

# Disable overseer
aiwg ralph-config set oversight.enabled false

# Change validation level
aiwg ralph-config set oversight.validation_level strict

ralph-config reset

Reset to default configuration.

aiwg ralph-config reset

Options:

--confirm - Skip confirmation prompt

ralph-config preset

Apply configuration preset.

aiwg ralph-config preset <name>

Arguments:

<name> - Preset name: conservative, balanced, aggressive

Presets:

Preset	Use Case	Settings
`conservative`	Security fixes, critical systems	Cautious gains, strict validation, strict oversight
`balanced`	General development (default)	Standard gains, standard validation, balanced oversight
`aggressive`	Documentation, rapid iteration	Aggressive gains, minimal validation, permissive oversight

Example:

# Set conservative preset for security work
aiwg ralph-config preset conservative

Capabilities: cli, ralph, configuration Platforms: All Tools: Read, Write

Documentation Commands

doc-sync

Synchronize documentation and code to eliminate drift.

aiwg doc-sync <direction> [options]

Arguments:

<direction> - Sync direction: code-to-docs, docs-to-code, full

Options:

--interactive - Prompt for each sync decision
--guidance "text" - Human guidance for ambiguous cases
--scope "path" - Limit to specific directory (default: .)
--dry-run - Audit only, no modifications
--parallel N - Max concurrent audit agents (default: 4)
--incremental - Git-diff since last sync instead of full scan
--verbose - Detailed per-file findings
--no-commit - Skip auto-commit
--max-iterations N - Ralph refinement iterations (default: 3)

Capabilities: cli, documentation, synchronization, audit Platforms: All Tools: Task, Read, Write, Bash, Glob, Grep, Edit

Directions:

Direction	Description
`code-to-docs`	Code is truth, update docs to match
`docs-to-code`	Docs are truth, generate TODOs/fixes for code
`full`	Bidirectional reconciliation

Execution phases:

Init and file inventory
Parallel domain audit (8 auditors)
Cross-reference validation
Drift report generation
Sync planning (auto-fixable / template-fixable / human-required)
Auto-fix application
Ralph refinement for complex items
Validation of changes
Record sync state and commit

Examples:

# Dry-run audit
aiwg doc-sync code-to-docs --dry-run

# Incremental sync after code changes
aiwg doc-sync code-to-docs --incremental --verbose

# Full bidirectional with guidance
aiwg doc-sync full --interactive --guidance "Focus on CLI reference"

# Scoped to specific directory
aiwg doc-sync code-to-docs --scope docs/extensions/

Output locations:

Audit report: .aiwg/reports/doc-sync-audit-{date}.md
Sync state: .aiwg/.last-doc-sync

SDLC Orchestration Commands

sdlc-accelerate

End-to-end SDLC ramp-up from idea to construction-ready.

aiwg sdlc-accelerate <description> [options]

Arguments:

<description> - Project description (idea entry point)

Options:

--from-codebase <path> - Scan existing code instead of starting from idea
--interactive - Full interactive mode at every step
--guidance "text" - Project-level guidance for all phases
--auto - Auto-proceed on CONDITIONAL gates
--dry-run - Show pipeline plan without executing
--skip-to <phase> - Jump to specific phase (validates prereqs)
--resume - Resume from detected current phase

Capabilities: cli, sdlc, orchestration, pipeline, accelerate Platforms: All Tools: Task, Read, Write, Glob, TodoWrite

Pipeline phases:

INTAKE → GATE_LOM → ELABORATION → GATE_ABM → CONSTRUCTION_PREP → BRIEF

Phase	Description	Delegates To
Intake	Project intake and inception	`/intake-wizard` or `/intake-from-codebase`
LOM Gate	Lifecycle Objective Milestone	`/flow-gate-check inception`
Elaboration	Architecture and requirements	`/flow-inception-to-elaboration`
ABM Gate	Architecture Baseline Milestone	`/flow-gate-check elaboration`
Construction Prep	Iteration planning	`/flow-elaboration-to-construction`
Brief	Construction Ready Brief	Template generation

Entry point detection:

Condition	Entry
No `.aiwg/` + description	`intake-wizard`
No `.aiwg/` + `--from-codebase`	`intake-from-codebase`
`.aiwg/` exists + `--resume`	Detect and resume
`--skip-to`	Jump with prereq validation

Examples:

# New project from idea
aiwg sdlc-accelerate "Customer portal with real-time chat"

# From existing codebase
aiwg sdlc-accelerate --from-codebase ./src "E-commerce platform"

# Resume interrupted pipeline
aiwg sdlc-accelerate --resume

# Preview pipeline plan
aiwg sdlc-accelerate --dry-run "Mobile banking app"

# Skip to elaboration
aiwg sdlc-accelerate --skip-to elaboration

# Auto-approve everything
aiwg sdlc-accelerate --auto "Quick prototype"

State tracking: .aiwg/reports/accelerate-state.json Output: .aiwg/reports/construction-ready-brief.md

Index Commands

Commands for building and querying the artifact index. The index provides structured, pre-computed metadata about project artifacts, enabling agents and developers to discover, search, and navigate artifacts without manual file searching.

The index uses a multi-graph architecture with three built-in graph types:

Graph	Scans	Storage	Built by default
`project`	`.aiwg/` artifacts	`.aiwg/.index/project/`	Yes
`codebase`	`src/`, `test/`, `tools/`	`.aiwg/.index/codebase/`	Yes
`framework`	`agentic/code/`, `docs/`	`.aiwg/.index/framework/`	No (use `--graph framework`)

All commands without --graph operate across all available project-local graphs (project + codebase). Use --graph <name> to target a specific graph.

index

Artifact index commands (build, query, deps, stats).

aiwg index <subcommand> [options]

Subcommands:

build - Build/rebuild the artifact index
query - Search artifacts by keyword, type, phase, tags
deps - Show artifact dependency graph
stats - Show index statistics

Global option (all subcommands):

--graph <type> - Target a specific graph: project, codebase, or framework

Capabilities: cli, index, artifacts, search, dependencies Platforms: All Tools: Read, Glob, Grep

index build

Build or rebuild the artifact index.

aiwg index build [options]

Options:

--force - Full rebuild (ignore checksums, re-index everything)
--verbose - Show detailed progress during indexing
--scope <dir> - Limit scan to a specific subdirectory
--graph <type> - Build a single graph only (project, codebase, framework)

Default behavior (no --graph): Builds project + codebase graphs only. The framework graph covers the AIWG framework source (agentic/code/, docs/) and must be built explicitly with --graph framework.

Incremental mode (default): Only re-indexes files whose checksum has changed. Use --force for a full rebuild.

Examples:

# Build project + codebase (default)
aiwg index build

# Full rebuild
aiwg index build --force

# Verbose output
aiwg index build --verbose

# Build framework graph (agentic/code/ + docs/)
aiwg index build --graph framework

# Build a single graph
aiwg index build --graph project

Output structure:

.aiwg/.index/
├── project/          # .aiwg/ artifacts
│   ├── metadata.json
│   ├── tags.json
│   ├── dependencies.json
│   └── stats.json
└── codebase/         # src/, test/, tools/
    ├── metadata.json
    ├── tags.json
    ├── dependencies.json
    └── stats.json

index query

Search artifacts by keyword, type, phase, tags, or path pattern.

aiwg index query [search-text] [options]

Arguments:

[search-text] - Optional keyword search (weighted: title 3x, tags 2x, summary 1x, path 0.5x)

Options:

--type <type> - Filter by artifact type (e.g., use-case, adr, test-plan)
--phase <phase> - Filter by SDLC phase (e.g., requirements, architecture, testing)
--tags <tag1,tag2> - Filter by tags (AND logic — all tags must match)
--path <glob> - Filter by file path glob pattern
--updated-after <date> - Filter by last-modified date
--limit <n> - Maximum number of results (default: 20)
--graph <type> - Search a specific graph only
--json - Output as JSON (recommended for agents)

Default behavior (no --graph): Searches across project + codebase graphs combined.

Examples:

# Search all project-local graphs
aiwg index query "authentication"

# Search framework source only
aiwg index query "artifact discovery" --graph framework

# Filter by type
aiwg index query --type use-case

# Combined filters
aiwg index query "login" --type use-case --phase requirements

# JSON output for agents
aiwg index query "auth" --json

index deps

Show artifact dependency graph based on @-mention references.

aiwg index deps <path> [options]

Arguments:

<path> - Path to the artifact (e.g., .aiwg/requirements/UC-001.md)

Options:

--direction <dir> - Direction: upstream, downstream, or both (default: both)
--depth <n> - Maximum traversal depth (default: 3)
--graph <type> - Use a specific graph's dependency data
--json - Output as JSON (recommended for agents)

Behavior:

upstream - What this artifact depends on (its @-mentions)
downstream - What depends on this artifact (mentions it)
both - Both directions

Default behavior (no --graph): Merges dependency data from project + codebase graphs.

Examples:

# Show all dependencies
aiwg index deps .aiwg/requirements/UC-001.md

# Downstream only (what would break if I change this?)
aiwg index deps .aiwg/requirements/UC-001.md --direction downstream

# JSON output with limited depth
aiwg index deps .aiwg/architecture/adr-001.md --depth 2 --json

# Deps within framework source
aiwg index deps agentic/code/frameworks/sdlc-complete/rules/artifact-discovery.md --graph framework

index stats

Show artifact index statistics and project health metrics.

aiwg index stats [options]

Options:

--graph <type> - Show stats for a specific graph only
--json - Output as JSON (recommended for agents)

Default behavior (no --graph):

Human-readable: shows each available graph with a section header
JSON: returns an object keyed by graph name with all stats

Reports:

Artifact counts by SDLC phase and type
Tag distribution
Dependency graph metrics (edges, orphaned artifacts)
Index coverage (indexed vs. total files)

Examples:

# Show all project-local graphs
aiwg index stats

# JSON output (aggregated, keyed by graph name)
aiwg index stats --json

# Single graph
aiwg index stats --graph project --json

# Framework graph stats
aiwg index stats --graph framework

Code Analysis Commands

cleanup-audit

Audit codebase for dead code, unused exports, orphaned files, and stale manifests.

aiwg cleanup-audit [--scope <path>] [--fix] [--verbose]

Capabilities: cli, analysis, code-quality, dead-code, cleanup Platforms: All Tools: Bash, Glob, Grep, Read, Write, Edit

Actions:

Scans for unused exports, orphaned files, and dead code
Detects stale manifest entries and broken references
Reports findings with severity classification
Optionally applies auto-fixes with --fix

Addon Commands

Commands contributed by installed addons. Available after running aiwg use <addon>.

Extension System

Unified Extension Schema

All commands are registered as extensions in the unified schema. This enables:

Dynamic discovery: Commands found via semantic search
Capability-based routing: Match commands by what they do
Auto-generated help: Help text always in sync
Platform awareness: Deploy to correct platform paths

Extension properties:

id: Unique identifier (kebab-case)
type: Extension type (command, agent, skill, etc.)
name: Human-readable name
description: Brief description
capabilities: What it can do
keywords: Search terms
platforms: Platform compatibility
metadata: Type-specific data

See also:

@src/extensions/types.ts - Full type definitions
@.aiwg/architecture/unified-extension-schema.md - Schema documentation

Command Categories

Category	Count	Commands
Maintenance	5	help, version, doctor, update, sync
Framework	3	use, list, remove
Project	1	new
Workspace	3	status, migrate-workspace, rollback-workspace
MCP	1	mcp (3 subcommands)
Catalog	1	catalog (3 subcommands)
Toolsmith	1	runtime-info
Utility	3	prefill-cards, contribute-start, validate-metadata
Plugin	5	install-plugin, uninstall-plugin, plugin-status, package-plugin, package-all-plugins
Scaffolding	8	add-agent, add-command, add-skill, add-behavior, add-template, scaffold-addon, scaffold-extension, scaffold-framework
Daemon	2	behavior, daemon-init
Ralph	8	ralph, ralph-status, ralph-abort, ralph-resume, ralph-attach, ralph-external, ralph-memory, ralph-config
Mission Control	1	mc (9 subcommands)
Agent Teams	1	team (3 subcommands)
Metrics	3	cost-report, cost-history, metrics-tokens
Documentation	1	doc-sync
SDLC Orchestration	1	sdlc-accelerate
Code Analysis	1	cleanup-audit
Index	1	index (4 subcommands)
Reproducibility	4	execution-mode, snapshot, checkpoint, reproducibility-validate
Addon: ring	5	ring check, ring circuit-breaker, ring session-start, ring session-end, ring status

Total: 53 built-in + addon commands (addon commands require aiwg use <addon>)

Exit Codes

Code	Meaning
0	Success
1	General error
2	Invalid arguments
3	Missing dependencies
4	Configuration error
5	Network error
10	Validation error
20	File system error

Environment Variables

Variable	Purpose	Default
`AIWG_HOME`	AIWG installation directory	Auto-detected
`AIWG_CHANNEL`	Update channel (stable/main)	`stable`
`AIWG_LOG_LEVEL`	Logging level (debug/info/warn/error)	`info`
`AIWG_USE_NEW_ROUTER`	Enable experimental router	`false`
`AIWG_LEGACY_MODE`	Force legacy routing	`false`

Configuration File

Optional .aiwgrc.json in project root:

{
  "defaultProvider": "claude",
  "autoUpdate": false,
  "frameworks": {
    "sdlc": {
      "agents": "all",
      "commands": ["use", "status", "help"]
    }
  },
  "teamProfile": {
    "project": "My Project",
    "team": "Platform Team",
    "defaultAuthor": "Developer Name"
  },
  "ralph": {
    "pid": {
      "enabled": true,
      "gain_profile": "standard"
    },
    "semantic_memory": {
      "enabled": true,
      "max_entries": 1000
    },
    "oversight": {
      "enabled": true,
      "intervention_mode": "balanced",
      "validation_level": "standard"
    }
  }
}

Common Workflows

Initial Setup

# Install globally
npm install -g aiwg

# Check installation
aiwg doctor

# Create new project
aiwg new my-project
cd my-project

Deploy to Existing Project

cd existing-project

# Deploy SDLC framework
aiwg use sdlc

# Check status
aiwg status

# Verify deployment
ls .claude/agents
ls .claude/commands

Multi-Platform Deployment

# Claude Code (default — auto-detected)
aiwg use sdlc

# GitHub Copilot
aiwg use sdlc --provider copilot

# Cursor
aiwg use sdlc --provider cursor

# Windsurf
aiwg use sdlc --provider windsurf

# Warp Terminal
aiwg use sdlc --provider warp

# Factory AI
aiwg use sdlc --provider factory

# OpenAI / Codex  (commands + skills deploy to ~/.codex/)
aiwg use sdlc --provider codex

# OpenCode
aiwg use sdlc --provider opencode

# Hermes (MCP sidecar — skills + lean AGENTS.md)
aiwg use sdlc --provider hermes

# OpenClaw (includes behaviors in ~/.openclaw/behaviors/)
aiwg use sdlc --provider openclaw

# Local / Ollama  (Claude Code paths, route coding tasks to local model)
aiwg use sdlc --provider local --coding-model ollama/qwen3.5:9b

# All platforms at once
aiwg use sdlc --provider all

Framework Management

# List installed
aiwg list

# Remove framework
aiwg remove marketing

# Reinstall with force
aiwg use marketing --force

Ralph Task Execution (Epic #26)

# Basic task
aiwg ralph "Fix failing tests" --completion "npm test passes"

# Security-critical with strict controls
aiwg ralph "Fix SQL injection" \
  --completion "security scan passes" \
  --gain-profile conservative \
  --validation-level strict \
  --intervention-mode strict

# Fast doc generation with minimal oversight
aiwg ralph "Update API docs" \
  --completion "docs/ updated" \
  --gain-profile aggressive \
  --disable-overseer

# Leverage past learnings
aiwg ralph "Optimize database queries" \
  --completion "benchmarks pass" \
  --enable-semantic-memory

# Check status mid-run
aiwg ralph-status

# Apply preset for common scenarios
aiwg ralph-config preset conservative
aiwg ralph "Migrate database" --completion "migration complete"

Troubleshooting

Command Not Found

# Check if installed globally
npm list -g aiwg

# Reinstall if missing
npm install -g aiwg

# Check PATH
echo $PATH

Permission Errors

# Fix npm permissions (Linux/Mac)
mkdir ~/.npm-global
npm config set prefix '~/.npm-global'
echo 'export PATH=~/.npm-global/bin:$PATH' >> ~/.bashrc
source ~/.bashrc

# Reinstall
npm install -g aiwg

Deployment Failures

# Run doctor
aiwg doctor

# Force reinstall
aiwg use sdlc --force

# Check logs
cat .aiwg/logs/deployment.log

MCP Issues

# Verify MCP server
aiwg mcp info

# Reinstall config
aiwg mcp install claude --force

# Test manually
aiwg mcp serve

Ralph Loop Issues (Epic #26)

# Check current status
aiwg ralph-status

# View configuration
aiwg ralph-config show

# Reset to defaults
aiwg ralph-config reset

# Inspect semantic memory
aiwg ralph-memory list

# Export state for debugging
aiwg ralph-memory export debug-memory.json

# Try different gain profile
aiwg ralph-config set pid.gain_profile conservative
aiwg ralph-resume

Support

Documentation: https://aiwg.io/docs
GitHub Issues: https://github.com/jmagly/aiwg/issues
Discord: https://discord.gg/BuAusFMxdA
Telegram: https://t.me/+oJg9w2lE6A5lOGFh

References

@src/extensions/commands/definitions.ts - All command definitions
@src/extensions/types.ts - Extension type system
@.aiwg/architecture/unified-extension-schema.md - Extension schema
@.aiwg/architecture/unified-extension-system-implementation-plan.md - Implementation details
@.aiwg/planning/epic-26-ralph-control-improvements.md - Epic #26 specification
@tools/ralph-external/ - Ralph external implementation
@.aiwg/ralph/ - Ralph state and memory storage
@CLAUDE.md - Project-level CLI integration
@README.md - Quick start guide

FilesExpand file tree

cli-reference.md

Latest commit

History

cli-reference.md

File metadata and controls

AIWG CLI Reference

Table of Contents

Maintenance Commands

help

version

doctor

update

sync

Framework Management

use

list

remove

Project Setup

new

Workspace Management

status

migrate-workspace

rollback-workspace

MCP Commands

mcp

mcp serve

mcp install

mcp info

Catalog Commands

catalog

catalog list

catalog info

catalog search

Toolsmith Commands

runtime-info

Schedule Skill

schedule create

schedule list

schedule delete

Backend Detection

Chrony Recommendation

Utility Commands

prefill-cards

contribute-start

validate-metadata

Plugin Commands

install-plugin

uninstall-plugin

plugin-status

package-plugin

package-all-plugins

Scaffolding Commands

Skills vs Commands — Provider Support

add-agent

add-command

add-skill

add-behavior

add-template

scaffold-addon

scaffold-extension

scaffold-framework

Daemon Commands

behavior

daemon-init

Mission Control Commands

mc

Agent Team Commands

team

Subcommands

Provider Routing

Options

Examples

Built-in Teams (sdlc-complete framework)

Team Definition Format

Ralph Commands

ralph

ralph-status

ralph-abort

ralph-resume