OriginRouter CLIExit status

Exit status and error recovery

Most CLI commands use 0 for success and 1 for a general failure. Collaboration commands provide a more specific stable exit-status set. Error prose is for people; automation should prioritize exit status and JSON fields.

Exit statuses are intended for scripts and terminal automation. For runtime state, approval requests, or device connectivity, use App Diagnostics and support, Approvals, and Devices and security.

General behavior

Status	Meaning
`0`	The command completed as requested. Still inspect JSON business state when required
`1`	General error, network error, invalid state, or a failure outside collaboration-specific classes
Child status	`originrouter run -- <command>` propagates the child process exit status
`143`	The Agent Workspace terminal UI received `SIGTERM`

An HTTP status is not a process exit status. Authentication and service failures are normally rendered as a readable headline, technical detail, and Next: action, then exit with 1.

Collaboration exit status

Status	Meaning	Recommended action
`2`	Invalid arguments, usage, draft, or input	Correct the command and inspect `originrouter --help`
`3`	Local daemon or Local API unavailable	Check `service status`, then `service restart` if needed
`4`	Login, device trust, E2EE, identity, or workspace permission failure	Check `login status`, trusted devices, and the permission profile
`5`	Runtime, provider, model, route, or capability unavailable	Run `doctor` on the target device and repair its route
`6`	The run entered `failed`	Run `collaboration doctor <run-id>`, fix the cause, then `retry`
`7`	Budget exhausted, budget exceeded, or policy blocked	Review run, device, agent budget, or organization policy
`8`	Configuration or plan requires confirmation in a non-interactive flow	Review the run and execute `confirm`; use `--yes` only when deliberately accepting both startup gates
`10`	Device offline, relay timeout, connection interruption, or incomplete Local API response	Inspect actual run state before deciding whether to create again

Status 10 can mean that the request result is uncertain. A daemon-owned run may still be active, so query the same run ID with show or doctor.

JSON automation

if output=$(originrouter collaboration show "$run_id" --json); then
  state=$(printf '%s' "$output" | jq -r '.state')
else
  status=$?
  printf 'originrouter failed with status %s\n' "$status" >&2
  exit "$status"
fi

For create commands, persist the first returned run ID. After a network interruption, query that ID; create another run only after confirming that the original request was not accepted.

Recovery order

Read the complete error, including Impact, Action, and Diagnostic code.
When supported, repeat a read-only command with --json.
Check login status, service status, and doctor.
For an existing collaboration, run collaboration doctor <run-id>.
Repeat a create or write operation only after confirming that the original request was not accepted.

Common authentication failures already map to concrete recovery actions such as signing in again, updating an old CLI, revoking an unused device, or waiting for the key-rotation limit to reset.