Command reference

Run commands from the project directory unless a command explicitly selects registered projects. This reference covers OpenWolf 2.5.2. Use openwolf --version and openwolf <command> --help to check your installed CLI.

Installation and status

openwolf init
openwolf init --agent claude codex opencode
openwolf status
openwolf status --all
openwolf operations doctor

init creates project memory and registers selected agents. status checks the installation. operations doctor also reports memory authority and update readiness. Existing malformed settings are preserved and reported for repair.

Project map

openwolf scan
openwolf scan --check
openwolf find validateToken
openwolf find --file src/auth.ts
openwolf map --focus auth,jwt --budget 1000

scan refreshes eligible index entries. scan --check checks freshness without writing and exits with a failure status when the index does not match. find searches indexed paths and symbols. map returns a focused overview within an estimated token budget.

Recorded usage

openwolf usage report --json
openwolf usage report --agent codex --json
openwolf usage reconcile
openwolf report

usage report reads available Claude, Codex and OpenCode records and reports coverage. usage reconcile persists the shared report and recovers retained observations. report also shows the operational ledger and estimates. None of these commands starts a model conversation.

Save a checkpoint

Create a JSON file with the task fields you want to save:

{
  "objective": "Fix expired-session handling",
  "next_action": "Run the authentication tests",
  "unresolved": ["The expired-session case is still failing"],
  "completed": ["Updated the token validation check"]
}

Use the actual receiving agent session ID:

openwolf handoff checkpoint --agent codex --session SESSION_ID --file checkpoint.json

Inspect and transfer a session

openwolf handoff list --from claude
openwolf handoff read --from claude --session SOURCE_ID
openwolf handoff export --from claude --session SOURCE_ID --to codex --preview
openwolf handoff export --from claude --session SOURCE_ID --to codex
openwolf handoff inspect PACKET_ID
openwolf handoff import PACKET_ID --to codex --session RECEIVING_ID

Replace the capitalised values with IDs from the current project. Inspect the preview before writing a packet, then inspect the packet before import. Import checks identity, source records and repository drift. --allow-drift is for explicitly reviewed historical evidence; it is not a repair for changed or missing source records.

openwolf handoff recover --from codex --session SESSION_ID
openwolf handoff search "expired session"

Recovery uses saved records when observations are missing. Search returns relevant saved evidence. See handover details.

Session notes and archives

openwolf memory log --session SESSION_ID --summary "Updated token validation" --files src/auth.ts --outcome "Expired-session test still fails"
openwolf memory archive --days 7 --dry-run
openwolf memory archive --days 7
openwolf memory restore ARCHIVE_ID
openwolf maintenance --dry-run

Archive previews do not write. Archival retains the latest block, pinned notes and tracked active sessions. Restore verifies the stored content before changing active memory. maintenance combines memory maintenance and usage reconciliation; inspect its preview first.

Durable memory review

openwolf memory review --json
openwolf operations prepare --output NEW_REVIEW_DIRECTORY

These commands produce review material. They do not grant authority. memory approve and memory revoke are independent administrator operations for a protected deployment. Follow the operations guide.

Dashboard and daemon

openwolf dashboard
openwolf daemon status
openwolf daemon start
openwolf daemon stop
openwolf daemon restart
openwolf daemon logs

The dashboard can start a local daemon without PM2. Persistent daemon management uses the supported PM2 integration where installed. Stop commands verify daemon ownership rather than stopping any process that happens to use a port.

Updates and backups

openwolf update --list
openwolf update --dry-run
openwolf update --project my-app
openwolf self-update --status
openwolf self-update
openwolf restore

update refreshes registered projects from the package already installed on your machine. self-update checks npm and prepares a compatible project runtime under the configured policy. It does not replace the global CLI. restore lists backups; restore BACKUP_NAME restores the selected snapshot and can replace newer project state.

Bugs and scheduled tasks

openwolf bug search "cannot read properties"
openwolf cron list
openwolf cron run TASK_ID
openwolf cron retry TASK_ID
openwolf cron enable TASK_ID
openwolf cron disable TASK_ID

Benchmarks

openwolf bench --repo /path/to/fixture --yes

The benchmark runs coding tasks with and without OpenWolf through the supported benchmark runner. It uses real model access and can consume paid usage. It requires --yes. Results apply to those tasks and runs, not every project or agent.