Skip to main content

​
openclaw automations

Manage automation jobs for the Gateway scheduler. openclaw automations is the primary command; openclaw cron remains an alias, and every subcommand below works with either spelling.
Run openclaw automations --help for the full command surface. See Automations for the conceptual guide.
All automation mutations (add/create, update/edit, remove, run) require operator.admin. Command-payload runs execute directly in the Gateway process, not as an agent tools.exec tool call; tools.exec.* and exec approvals still govern model-visible exec tools.
Every automation subcommand accepts the shared Gateway connection options. Use --port <port> for a Gateway on a non-default local port, or --url <url> for an explicit WebSocket URL; do not combine them. Connection options such as --port, --url, and --token may appear before or after the subcommand.

​
Create jobs quickly

openclaw automations create is an alias for openclaw automations add. For new jobs, put the schedule first and the prompt second:
Use --webhook <url> when the job should POST the finished payload instead of delivering to a chat target:
Use --command for deterministic shell-style jobs that run inside the OpenClaw scheduler without starting an isolated agent/model run:
--command <shell> stores argv: ["sh", "-lc", <shell>]. Use --command-argv '["node","scripts/report.mjs"]' for exact argv execution. Command jobs capture stdout/stderr, record normal run history, and route output through the same announce, webhook, or none delivery modes as isolated jobs. A command that prints only NO_REPLY is suppressed. Use --display-name <name> when the list and detail views should show a human-readable label distinct from the automation’s stable name. Set or update that label with automations add|edit --display-name. Use automations edit <job-id> --clear-display-name to remove the label and restore the stable name in list and detail views. The set and clear options cannot be combined.

​
Sessions

--session accepts main, isolated, current, or session:<id>. Agent-turn jobs default to the creating conversation when session context is available. Without a session key, including ordinary CLI calls and API calls that omit one, the target falls back to isolated.
  • main binds to the agent’s main session.
  • isolated creates a fresh transcript and session id for each run.
  • current binds to the active session at creation time.
  • session:<id> pins to an explicit persistent session key.
Isolated runs reset ambient conversation context. Channel and group routing, send/queue policy, elevation, origin, and ACP runtime binding are reset for the new run. Safe preferences and explicit user-selected model or auth overrides can carry across runs.

​
Delivery

openclaw automations list and openclaw automations show <job-id> preview the resolved delivery route. For channel: "last", the preview shows whether the route resolved from the main or current session, or will fail closed. Provider-prefixed targets can disambiguate unresolved announce channels. For example, to: "telegram:123" selects Telegram when delivery.channel is omitted or last. Only prefixes advertised by the loaded plugin are provider selectors. If delivery.channel is explicit, the prefix must match that channel; channel: "whatsapp" with to: "telegram:123" is rejected. Service prefixes such as imessage: and sms: remain channel-owned target syntax.
Isolated automations add jobs default to --announce delivery. Use --no-deliver to keep output internal. --deliver remains as a deprecated alias for --announce.

​
Delivery ownership

Isolated automation chat delivery is shared between the agent and the runner:
  • The agent can send directly using the message tool when a chat route is available.
  • announce fallback-delivers the final reply only when the agent did not send directly to the resolved target.
  • webhook posts the finished payload to a URL.
  • none disables runner fallback delivery.
Use automations add|create --webhook <url> or automations edit <job-id> --webhook <url> to set webhook delivery. Do not combine --webhook with chat delivery flags such as --announce, --no-deliver, --channel, --to, --thread-id, or --account. automations edit <job-id> can unset individual delivery routing fields with --clear-channel, --clear-to, --clear-thread-id, and --clear-account (each is rejected when combined with its matching set flag). Unlike --no-deliver, which only disables runner fallback delivery, these remove the stored field so the job resolves that part of its route from defaults again. --announce is runner fallback delivery for the final reply. --no-deliver disables that fallback but does not remove the agent’s message tool when a chat route is available. Reminders created from an active chat preserve the live chat delivery target for fallback announce delivery. Internal session keys may be lowercase; do not use them as a source of truth for case-sensitive provider IDs such as Matrix room IDs.

​
Failure delivery

Failure notifications resolve in this order:
  1. Route fields in the job’s failureAlert object.
  2. delivery.failureDestination on the job, layered over the global destination fields on cron.failureAlert (mode, channel, to, accountId). The retired cron.failureDestination block is merged into them by openclaw doctor --fix.
  3. The job’s primary announce target (when neither of the above resolves to a concrete destination).
Jobs with one of those routes default to an execution-failure alert after 2 consecutive failures and a 1-hour cooldown. A per-job or global failureAlert object explicitly activates/tunes the policy even without an existing route. failureAlert: false disables execution and required-delivery failure alerts for the job, but not the auto-disable safety notification. Global enabled: false disables inheritance unless the job has its own failureAlert object. delivery.bestEffort: true suppresses inherited/default execution alerts, but not an explicit per-job policy.
Main-session jobs may only use delivery.failureDestination when primary delivery mode is webhook. Isolated jobs accept it in all modes.
Chat failure notifications include the run start time in the agent’s configured user timezone. Webhook message text stays stable and exposes the instant as runAtMs. Isolated automation runs treat run-level agent failures as job errors even when no reply payload is produced, so model/provider failures still increment error counters and trigger failure notifications. Command jobs do not start an isolated agent turn. A zero exit code records ok; non-zero exit, signal, timeout, or no-output timeout records error and can trigger the same failure notification path. Required completion delivery is separate: status: "ok" with completionStatus: "failed" does not increment the execution streak or backoff. It can notify immediately only through a resolved alternate failure destination, never the primary route that just failed. If an isolated run times out before the first model request, openclaw automations show and openclaw automations runs include a phase-specific error such as setup timed out before runner start or a stall message naming the last-known startup phase (for example context-engine). For CLI-backed providers, the pre-model watchdog stays active until the external CLI turn starts, so session lookup, hook, auth, prompt, and CLI setup stalls are reported as pre-model automation failures.

​
Scheduling

​
One-shot jobs

--at <datetime> schedules a one-shot run. Offset-less datetimes are treated as UTC unless you also pass --tz <iana>, which interprets the wall-clock time in the given timezone.
One-shot jobs delete only after completionStatus: "succeeded". Required-delivery failure or unknown completion keeps the job disabled, with no next run, so restarts do not replay payload side effects. Use --keep-after-run to preserve successful jobs too.

​
Recurring jobs

Recurring jobs use exponential retry backoff after consecutive errors: 30s, 1m, 5m, 15m, 60m. The schedule returns to normal after the next successful run. Skipped runs are tracked separately from execution errors. They do not affect retry backoff, but openclaw automations edit <job-id> --failure-alert-include-skipped can opt failure alerts into repeated skipped-run notifications. For isolated jobs that target a local configured model provider (base URL on loopback, a private network, or .local), the scheduler runs a lightweight provider preflight before starting the agent turn: api: "ollama" providers are probed at /api/tags; other local OpenAI-compatible providers (api: "openai-completions", e.g. vLLM, SGLang, LM Studio) are probed at /models. If the endpoint is unreachable, the run is recorded as skipped and retried on a later schedule; the reachability result is cached per endpoint for 5 minutes so many jobs against the same local server do not hammer it with repeated probes. Automation jobs, pending runtime state, and run history live in the shared SQLite state database. Legacy jobs.json, <name>-state.json, and runs/*.jsonl files are imported once and renamed with a .migrated suffix. After import, edit schedules with openclaw automations add|edit|remove instead of editing JSON files.

​
Manual runs

openclaw automations run <job-id> force-runs by default and returns as soon as the manual run is queued. Successful responses include { ok: true, enqueued: true, runId }. Use the returned runId to inspect the later result:
Add --wait when a script should block until that exact queued run records a terminal status:
With --wait, the CLI calls cron.run first, then polls the durable cron.runs row for the returned runId; it does not reread mutable job delivery settings. JSON reports payload execution as status and whole-run completion as completionStatus. The command exits 0 only for completionStatus: "succeeded"; failed, unknown, execution errors/skips, a missing runId, and timeout expiry exit non-zero (default 10m, polled every 2s by default). --poll-interval must be greater than zero.
Use --due when you want the manual command to run only if the job is currently due. If --due --wait does not enqueue a run, the command returns the normal non-run response instead of polling.

​
Models

automations add|edit --model <ref> selects an allowed model for the job. automations add|edit --fallbacks <list> sets per-job fallback models, for example --fallbacks openrouter/gpt-4.1-mini,openai/gpt-5; pass --fallbacks "" for a strict run with no fallbacks. automations edit <job-id> --clear-fallbacks removes the per-job fallback override. automations edit <job-id> --clear-model removes the per-job model override so the job follows normal automation model-selection precedence (a stored automation-session override if present, otherwise the agent/default model); it cannot be combined with --model. automations add|edit --thinking <level> sets a per-job thinking override; automations edit <job-id> --clear-thinking removes it so the job follows normal automation thinking precedence, and it cannot be combined with --thinking.
If the model is not allowed or cannot be resolved, the scheduler fails the run with an explicit validation error instead of falling back to the job’s agent or default model selection.
The automation --model is a job primary, not a chat-session /model override. That means:
  • Configured model fallbacks still apply when the selected job model fails.
  • Per-job payload fallbacks replaces the configured fallback list when present.
  • An empty per-job fallback list (--fallbacks "" or fallbacks: [] in the job payload/API) makes the run strict.
  • When a job has --model but no fallback list is configured, OpenClaw passes an explicit empty fallback override so the agent primary is not appended as a hidden retry target.
  • Local-provider preflight checks walk configured fallbacks before marking a run skipped.
openclaw doctor reports jobs that already have payload.model set, including provider namespace counts and mismatches against agents.defaults.model. Use that check when auth, provider, or billing behavior looks different between live chat and scheduled jobs.

​
Isolated automation model precedence

Isolated automation runs resolve the active model in this order:
  1. Gmail-hook override.
  2. Per-job --model.
  3. Stored automation-session model override (when the user selected one).
  4. Agent or default model selection.

​
Fast mode

Isolated automation fast mode follows the resolved live model selection. It resolves stored session fastMode, per-agent agents.entries.*.fastModeDefault, global agents.defaults.fastModeDefault, then selected-model params.fastMode. When the resolved mode is auto, the cutoff uses the selected model’s params.fastAutoOnSeconds value, defaulting to 60 seconds.

​
Live model switch retries

If an isolated run throws LiveSessionModelSwitchError, the scheduler persists the switched provider and model (and switched auth profile override when present) for the active run before retrying. The outer retry loop is bounded to two switch retries after the initial attempt, then aborts instead of looping forever.

​
Run output and denials

​
Stale acknowledgement suppression

Isolated automation turns suppress stale acknowledgement-only replies. If the first result is just an interim status update and no descendant subagent run is responsible for the eventual answer, the scheduler re-prompts once for the real result before delivery.

​
Silent token suppression

If an isolated automation run returns only the silent token (NO_REPLY or no_reply), the scheduler suppresses both direct outbound delivery and the fallback queued summary path, so nothing is posted back to chat.

​
Structured denials

Isolated automation runs use structured execution-denial metadata from the embedded run (fatal exec-tool errors coded SYSTEM_RUN_DENIED or INVALID_REQUEST) as the authoritative denial signal. They also honor node-host UNAVAILABLE wrappers around a nested structured error carrying one of those codes. The scheduler does not classify final-output prose or approval-looking refusal phrases as denials unless the embedded run also provides structured denial metadata, so ordinary assistant text is not treated as a blocked command. automations list and run history surface the denial reason instead of reporting a blocked command as ok.

​
Retention

Retention behavior:
  • cron.sessionRetention (default 24h, or false to disable; a zero duration such as "0h" also disables) prunes completed isolated run sessions.
  • Terminal run history is retained for 7 days (lost rows for 24 hours), with the newest 2000 rows per job and history class enforced as an additional ceiling.

​
Migrating older jobs

If you have automation jobs from before the current delivery and store format, run openclaw doctor --fix. Doctor normalizes legacy job fields (jobId, schedule.cron, top-level delivery fields including legacy threadId, payload provider delivery aliases) and migrates notify: true webhook fallback jobs from the retired raw cron.webhook value to explicit webhook delivery before removing that config key. Jobs that already announce to a chat keep that delivery and get a completion webhook destination. Without a legacy webhook, the inert top-level notify marker is removed for jobs with no migration target (the existing delivery is preserved unchanged), so doctor --fix no longer keeps re-warning about them.

​
Common edits

Update delivery settings without changing the message:
Disable delivery for an isolated job:
Enable lightweight bootstrap context for an isolated job:
Announce to a specific channel:
Announce to a Telegram forum topic:
Create an isolated job with lightweight bootstrap context:
--light-context applies to isolated agent-turn jobs only. For automation runs, lightweight mode keeps bootstrap context empty instead of injecting the full workspace bootstrap set. Create a command job with exact argv, cwd, env, stdin, and output limits:

​
Common admin commands

Manual run and inspection:
openclaw automations list shows enabled jobs by default. Pass --all to include disabled jobs, or --agent <id> to show only jobs whose effective normalized agent id matches; jobs without a stored agent id count as the configured default agent. --json always requests JSON output. Commands whose product is already a machine-readable result emit JSON results by default: add/create, status, enable, disable, rm/remove/delete, run, edit, get, and runs. They accept --json as the explicit machine-output spelling. openclaw automations get <job-id> returns the stored job JSON directly; use automations show <job-id> when you want the human-readable view with delivery-route preview. list and show use human-readable output by default and switch to JSON with --json. scratch reads raw scratch content by default and prints the scratch plus revision metadata with --json; scratch writes return the revision result as JSON by default and accept --json as the explicit machine-output spelling. automations list --json and automations show <job-id> --json include a top-level status field on each job, computed from enabled, state.runningAtMs, and state.lastRunStatus. Values: disabled, running, ok, error, skipped, or idle. JSON status stays canonical and undecorated so external tooling can read job state without re-deriving it; human output may decorate repeated error statuses with a failure count. automations runs entries include delivery diagnostics with the intended automation target, the resolved target, message-tool sends, fallback use, and delivered state. Private per-job scratch (heartbeat checklists and similar monitor context):
Scratch is stored in the shared state database, capped at 256 KiB, and never included in automations list/automations get/automations runs output. Writes are compare-and-swap guarded against the revision read at command start; pass --expected-revision <n> to pin an explicit revision instead. See Heartbeat for how heartbeat monitors use scratch. Agent and session retargeting:
openclaw automations add warns when --agent is omitted on agent-turn jobs and falls back to the default agent (main). Pass --agent <id> at create time to pin a specific agent. Delivery tweaks: