Commands

Complete reference for all OpenWolf CLI commands.

openwolf init

Initialize OpenWolf in the current project.

openwolf init

What it does:

1. Detects the project root (looks for .git, package.json, Cargo.toml, etc.)
2. Creates .wolf/ with 10 template files
3. Copies 7 hook scripts to .wolf/hooks/
4. Registers 6 Claude Code hooks in .claude/settings.json
5. Creates .claude/rules/openwolf.md
6. Prepends the OpenWolf snippet to CLAUDE.md
7. Runs an initial anatomy scan
8. Populates cerebrum.md with detected project name and description

If .wolf/ already exists, it reinitializes (overwrites templates, preserves learned data).

INFO

If .claude/settings.json already has hooks, OpenWolf merges its hooks in -- existing hooks are not overwritten.

---

openwolf status

Show health, stats, and file integrity.

openwolf status

Output:

OpenWolf Status
===============

  ✓ All 10 core files present
  ✓ All 7 hook scripts present
  ✓ Claude Code hooks registered (6 matchers)

Token Stats:
  Sessions: 12
  Total reads: 87
  Total writes: 34
  Tokens tracked: ~45,200
  Estimated savings: ~8,400 tokens

Anatomy: 79 files tracked

Daemon: running
  Last heartbeat: 2 minutes ago

---

openwolf scan

Force a full anatomy rescan of the project.

openwolf scan

Scanning project...
  ✓ Anatomy scan complete: 79 files indexed in 42ms

openwolf scan --check

Compare the current filesystem against anatomy.md without writing any changes. Exits with code 1 if the anatomy is out of date.

openwolf scan --check

Useful in CI pipelines to verify that anatomy.md has been kept in sync:

openwolf scan --check || echo "anatomy.md is out of date. Run openwolf scan"

---

openwolf dashboard

Open the real-time dashboard in your default browser.

openwolf dashboard

If the daemon is not running, this command automatically starts it as a background process. No PM2 required.

The dashboard opens at http://localhost:18791 and connects via WebSocket for live updates.

---

openwolf daemon

Manage the background daemon process.

openwolf daemon start

Start the daemon via PM2 for persistent background operation.

openwolf daemon start

The daemon handles:

• Cron tasks (anatomy rescans, memory consolidation, AI reflections)
• File watching and WebSocket broadcasting
• Dashboard HTTP server
• Health heartbeat

PM2 is optional

You do not need PM2 to use the daemon. Running openwolf dashboard starts the daemon automatically via Node's fork(). PM2 is only needed for auto-restart and boot persistence.

Windows

Run pm2-windows-startup for boot persistence.

openwolf daemon stop

Stop the running daemon. Works whether the daemon was started via PM2 or via openwolf dashboard:

openwolf daemon stop

• First tries to stop via PM2
• Falls back to finding the process listening on the dashboard port and killing it directly

openwolf daemon restart

openwolf daemon restart

openwolf daemon logs

Show the last 50 lines of daemon output.

openwolf daemon logs

---

openwolf cron

Manage scheduled cron tasks.

openwolf cron list

Show all tasks with their schedule, status, and last run time.

openwolf cron list

Cron Tasks
==========

  Full anatomy rescan (anatomy-rescan)
    Schedule: Every 6 hours
    Status: enabled
    Last run: 3 hours ago

  Consolidate old memory (memory-consolidation)
    Schedule: Daily at 2:00 AM
    Status: enabled
    Last run: yesterday

  Token audit report (token-audit)
    Schedule: Mondays at midnight
    Status: enabled
    Last run: 5 days ago

  Cerebrum reflection (cerebrum-reflection)
    Schedule: Sundays at 3:00 AM
    Status: enabled
    Last run: 2 days ago
    Uses: claude -p (subscription)

  AI suggestions (project-suggestions)
    Schedule: Mondays at 4:00 AM
    Status: enabled
    Last run: 5 days ago
    Uses: claude -p (subscription)

openwolf cron run <id>

Manually trigger a cron task by ID.

openwolf cron run anatomy-rescan

openwolf cron run project-suggestions

The command first attempts to dispatch the task through the daemon's HTTP API. If the daemon is not running, it falls back to executing the task directly in the current process. The daemon is not required.

openwolf cron retry <id>

Remove a task from the dead letter queue so it retries on its next schedule.

openwolf cron retry cerebrum-reflection

---

openwolf update

Update all registered OpenWolf projects to the latest templates.

openwolf update

Options:

Flag	Description
--dry-run	Show what would be updated without making changes
--project <name>	Update only a specific project (partial name match)
--list	List all registered projects and their paths

Before making any changes, update creates a timestamped backup of the existing .wolf/ directory. You can restore from these backups with openwolf restore.

# Preview changes without writing anything
openwolf update --dry-run

# Update a single project
openwolf update --project my-app

# See all registered projects
openwolf update --list

---

openwolf restore

Restore .wolf/ from a backup created by openwolf update. Run this from the project directory.

openwolf restore [backup]

Without arguments, lists all available backups:

openwolf restore

With a backup name, restores from that backup:

openwolf restore 2026-03-15T10-30-00

---

openwolf designqc

Capture full-page screenshots for design evaluation by Claude Code.

openwolf designqc [target]

Options:

Flag	Default	Description
--url <url>	(auto-detect)	Dev server URL. Auto-starts the dev server if omitted
--routes <routes...>	(all detected)	Specific routes to capture
--quality <n>	70	JPEG quality 1--100. Lower values produce smaller files and consume fewer tokens
--max-width <n>	1200	Maximum capture width in pixels
--desktop-only	false	Skip mobile viewport captures

Requires: puppeteer-core must be installed in the project or globally.

npm install puppeteer-core

How it works:

1. Detects or starts the project's dev server (Vite, Next.js, etc.)
2. Launches a headless Chrome/Edge instance
3. Captures full-page screenshots, split into sections (max 8 sections per route) for large pages
4. Saves all captures to .wolf/designqc-captures/

Screenshots are JPEG at the configured quality to keep file sizes and token counts low.

Full workflow:

# Step 1: capture screenshots
openwolf designqc

# Step 2: ask Claude to evaluate the design
# In your Claude Code session, say:
#   "Read the screenshots in .wolf/designqc-captures/ and evaluate the design"

Claude evaluates the screenshots inline using its vision capabilities. No separate API key is needed beyond your existing Claude Code subscription.

Examples:

# Capture specific routes only
openwolf designqc --routes / /about /pricing

# Point at a running server, desktop only
openwolf designqc --url http://localhost:3000 --desktop-only

# Lower quality for faster iteration
openwolf designqc --quality 40 --max-width 800

---

openwolf bug search <term>

Search the bug memory for matching entries.

openwolf bug search "cannot read properties"

Found 2 matching bug(s):

  [bug-003] TypeError: Cannot read properties of undefined (reading 'map')
    File: src/components/UserList.tsx
    Root cause: API response was null when users array was expected
    Fix: Added optional chaining: data?.users?.map() and fallback empty array
    Tags: null-check, api-response, typescript, react
    Occurrences: 3 | Last seen: 2026-03-09T14:30:00Z

Searches across: error messages, root causes, fixes, tags, and file paths.

---

openwolf --version

Print the current OpenWolf version. The version is read from package.json at runtime.

openwolf --version

1.0.0