OpenClawDocs
On this page

On this page

View as MarkdownView this page as plain textOpen in ChatGPTAsk questions about this pageOpen in ClaudeAsk questions about this page

CLI commands

Health

openclaw health

Fetch a health snapshot from the running Gateway over WebSocket RPC (no direct channel sockets from the CLI).

Options

Flag Default Description
--json false Print machine-readable JSON instead of text.
--timeout <ms> 60000 Connection timeout in milliseconds.
--verbose false Forces a live check and expands output across all configured accounts and agents.
--debug false Alias for --verbose.

Examples:

bash
openclaw healthopenclaw health --jsonopenclaw health --timeout 2500openclaw health --verboseopenclaw health --debug

Behavior

  • Local Gateway startup uses the shared 60-second readiness budget and reports its observed phase. An explicit --timeout overrides that budget. Startup still in progress at the deadline returns a non-failing result; JSON reports { "status": "starting", "startupPhase": "…" } instead of a health snapshot.

  • Local readiness recognizes wildcard listeners that serve loopback, including on macOS where a separate loopback bind can succeed on the same port.

  • Without --verbose, the Gateway can return a cached snapshot and refresh it in the background for the next caller. A cached snapshot is fresh for up to 60 seconds and unchanged from live channel runtime state.

  • --verbose forces a live check of each channel account. It also prints Gateway connection details. It expands human-readable output across all configured accounts and agents, instead of just the default agent.

  • An unconfigured or disabled preferred account does not hide check results from other active accounts. Ordinary output still follows the default agent's account bindings. --verbose includes all accounts.

  • When the displayed account is disabled, health text shows disabled and status --deep marks it OFF, even if the account remains configured.

  • Unhealthy channel lines include the recorded startup error when available, so a stopped channel reports its failure cause alongside its state.

  • Human-readable output includes failures for plugins enabled explicitly, automatically, or by default, and warnings for configured plugins that are unavailable. It shows at most 20 plugin diagnostics plus an omitted count. These warnings also appear in openclaw gateway health and the Health table in openclaw status --deep. Plugin diagnostics are sanitized for single-line terminal output; JSON retains the snapshot values.

  • Once ready, --json returns the full snapshot: channels, per-account checks, plugin load state, context-engine quarantine state, model-pricing cache state, event-loop health, delivery-queue warnings, and per-agent session stores.

  • Config read failures report the unreadable path and underlying error instead of a missing-credentials diagnostic. This also applies to openclaw gateway health.

  • Session ages in text and JSON use the Gateway's clock.

  • Heartbeat intervals in text show the resolved cadence without rounding away milliseconds. Week units are retained for long intervals.

  • Top-level ok: true means the health RPC succeeded and the Gateway produced a snapshot. Queue and plugin warnings do not change it to false.

  • When outbound or session deliveries, or inbound channel events, are dead-lettered, text output reports their counts and oldest failure age. Inbound counts are grouped by channel account. Inspect or recover individual events with openclaw channels dead-letters.

  • Optional deliveryQueues.ingressPressure summarizes durable inbound lanes that may be blocking later events. It is grouped by channel account and never exposes event, lane, payload, error, owner, token, session, or target identifiers. See Gateway health for the exact qualification and counting semantics.

Was this useful?