CLI Reference
Updated at 2026-09-21 · 623d849
The digital-me CLI installs, diagnoses, and orchestrates the ecosystem.
Run any command with --help for this text — it is extracted verbatim
from the CLI source on every release, so it cannot go stale.
Commands
| Command | What it does |
|---|---|
setup | One-shot: detect CLIs, init wiki dir, install brain-host (the hub: memory_search + orchestrator on :18791, scheduler on), install the adapters pointed at it, link the digital-me command onto PATH, doctor. |
init | Scaffold wiki/, inbox/, .cache/, config.example.yaml. |
doctor | Diagnose the environment. |
install | --runtime brain-host links the brain host, writes its token file, builds the retrieval index (GEMINI_API_KEY from /.data/.env, or the legacy ~/.openclaw/.env) and installs the always-on service. |
dashboard | Launch the OA dashboard: open it in your browser if it's already serving; otherwise start the always-on service first. |
service | Manage the always-on dashboard service (cross-platform: launchd on macOS, systemd --user on Linux). |
update | Update openclaw to the latest mature stable upstream tag and re-materialize the digital-me plugin overlay (stock + overlay model: no fork, no rebase). |
deploy | Turn 'merged in git' into 'verified live' in one safe step: refuse if the source checkout is dirty/ahead, fast-forward main, rebuild, reinstall to the canonical loaded location, restart the service, and self-verify the live fingerprint matches. |
migrate | One-shot copy from upstream task-orchestrator.db to brain.db. |
brain-db | Move brain.db from the legacy ~/.openclaw/data/ location to its canonical home /.data/brain.db (SQLite VACUUM INTO — a consistent snapshot). |
dream-cycle | Run the dream-cycle knowledge distillation pipeline. |
Full usage
digital-me — install, doctor, and orchestrate the digital-me ecosystem.
Usage:
digital-me setup [--wiki-root <path>] [--minimal]
One-shot: detect CLIs, init wiki dir, install brain-host (the hub:
memory_search + orchestrator on :18791, scheduler on), install the
adapters pointed at it, link the `digital-me` command onto PATH,
doctor. openclaw is optional — its plugin is installed only where
openclaw is present. --minimal skips the heavy optional services
(dream-cycle Python venv + dashboard build + digest).
digital-me init [--wiki-root <path>]
Scaffold wiki/, inbox/, .cache/, config.example.yaml.
digital-me doctor [--runtime <id>...]
Diagnose the environment.
digital-me install --runtime <id> [--runtime <id>...] [--no-service]
--runtime brain-host links the brain host, writes its token file,
builds the retrieval index (GEMINI_API_KEY from <wiki-root>/.data/.env,
or the legacy ~/.openclaw/.env) and installs the always-on service.
DIGITAL_ME_BRAIN_SCHEDULER=on|off sets its scheduler tick (default on
when no other host ticks brain.db). Adapters need a brain endpoint:
brain-host (installed first when listed) or a legacy openclaw gateway.
Install a specific runtime adapter. For --runtime dashboard, also
sets up an always-on service (launchd/systemd) so the dashboard
survives closing the terminal + reboot; --no-service skips that.
digital-me dashboard [--port <n>] [--no-open]
Launch the OA dashboard: open it in your browser if it's already
serving; otherwise start the always-on service first. --no-open
prints the URL instead of opening a browser.
digital-me service dashboard <install|uninstall|status>
Manage the always-on dashboard service (cross-platform: launchd on
macOS, systemd --user on Linux). 'install' generates + loads the unit
anchored at the stable install symlink and verifies it serves.
digital-me update --runtime openclaw [--dry-run] [--skip-restart]
[--repo-dir <path>] [--tag-maturity-hours <n>] [--pnpm-spec <spec>]
Update openclaw to the latest mature stable upstream tag and
re-materialize the digital-me plugin overlay (stock + overlay model:
no fork, no rebase). --dry-run prints the plan without writing;
--skip-restart leaves the gateway untouched.
digital-me deploy [--runtime openclaw|dashboard ...] [--dry-run]
Turn 'merged in git' into 'verified live' in one safe step: refuse if
the source checkout is dirty/ahead, fast-forward main, rebuild, reinstall
to the canonical loaded location, restart the service, and self-verify the
live fingerprint matches. No --runtime → every deployable runtime detected.
digital-me migrate [--from <path>] [--to <path>]
One-shot copy from upstream task-orchestrator.db to brain.db.
Idempotent — re-runs skip already-migrated rows.
digital-me brain-db migrate [--dry-run]
Move brain.db from the legacy ~/.openclaw/data/ location to its
canonical home <wiki-root>/.data/brain.db (SQLite VACUUM INTO — a
consistent snapshot). Refuses while brain-host is serving it; prints
the re-install steps every reader needs afterwards.
digital-me dream-cycle [args...]
Run the dream-cycle knowledge distillation pipeline. All args
after `dream-cycle` pass through to `python3 -m dream_cycle.run`.
digital-me dream-cycle import-workflow <path>
Import a workflow.json into the brain (brain-host, or a legacy gateway).
digital-me dream-cycle --via-agents [args...]
Run dream-cycle via brain-orchestrator spawn-dispatch instead of
inline-Python LLM. Requires an imported workflow + a running brain.
Runtimes:
claude-code 5 hooks + digital-me skill into ~/.claude/
codex CODEX.md + openclaw-brain MCP into ~/.codex/
hermes SOUL.md digital-me protocol into ~/.hermes/
openclaw (optional) gateway plugin — see @digital-me/runtime-openclaw README
brain-host the hub: retriever + orchestrator service on :18791 (installed by setup)
dream-cycle creates ~/.venvs/dream-cycle/ + pip install -e the Python service
dashboard builds the dashboard + installs the always-on service (see --no-service)