Skip to content

Workflow commands ​

The generated commands map the REST API one to one. Six top-level workflow commands combine several requests for the questions a developer, an operator or an agent asks most. They select a process instance by id, --business-key or --process-definition-key <key> --latest, and their views list ready-to-run follow-ups under next.

This page is the overview. The CLI has the full text: operate guide workflow for the options and rules, operate describe <command> for the requests a command sends.

CommandEffectWhat it does
operate inspectreadWhere an instance waits and why: wait states, root incidents, variables
operate waitread (write with --execute-jobs)Waits until the instance is idle or reaches an --until condition
operate advancewriteCompletes what the instance waits for
operate retrywrite (bulk for a whole definition)Sets the retries behind open incidents; --now executes the jobs
operate deploywriteDeploys changed files and directories, can start an instance
operate statusreadEngine triage: incidents by root cause, overdue jobs, workers, batches

The development loop, without a single id passed around:

sh
operate deploy bpmn --start --business-key B-1 --var amount=250
operate inspect --business-key B-1
operate advance --business-key B-1 --var approved=true --wait
operate retry --business-key B-1 --now
operate wait --business-key B-1 --until ended
operate status

On a terminal, or with -o table, a view prints as sections:

text
$ operate inspect --business-key ORD-1001 -o table
Process instance  391a7907-c652-11f1-9fc4-1a2aceba9a9d
State             ACTIVE
Definition        order-process v1 (Order process)
Business key      ORD-1001
Started           2026-10-12T15:33:07.285+0000

Waiting at:
ACTIVITY      KIND      ID                                    DETAIL        INSTANCE
review-order  userTask  391a790c-c652-11f1-9fc4-1a2aceba9a9d  Review order

Variables:
NAME    VALUE  TYPE
amount  250

Next:
  operate advance 391a7907-c652-11f1-9fc4-1a2aceba9a9d
  • Waiting replaces sleep and poll loops: operate wait, and --wait / --until of advance, retry and deploy, poll until the condition holds or --wait-timeout (default 60s) runs out. An incident ends the wait at once.
  • Exit code 9: every request succeeded, but the process did not reach the expected state (WAIT_TIMEOUT, INCIDENT, INSTANCE_ENDED, JOB_FAILED, CHECK_FAILED). The command still prints its view on stdout.
  • --dry-run: inspect, wait and status send nothing; advance and retry send their reads and preview the writes; deploy previews the upload.
  • Guards: inspect, wait and status work in read-only mode; advance, deploy, retry and wait --execute-jobs are writes; retry --process-definition-key alone is a bulk operation that needs --yes.
  • advance needs --activity-id when the instance waits at several places; the error lists them with ready commands.