get

graphenectl get all|<kind> [flags]        # list
graphenectl get <kind> <id> [flags]       # one record

get lists records of a kind — or all of them — and reads one record in full: dimension 1 of the five, the state. A run is a kind like any other (get run); the listing then shows run columns.

Flags

Flag	Type	Default	What it does
-l, --selector k=v	repeatable	—	label selector, every pair must match
-p, --phase <word>	string	—	THE lifecycle filter, one vocabulary for everything: a record's creating, ready, deleting, deleted; a run's running, completed, failed, canceled, terminated, timed-out
--owner <ref>	string	—	records owned by this owner (run/x, stand/p, agent/vm-1)
-w, --watch	bool	off	watch: the snapshot, then only changes — see Output forms
--chunk-size	int	500	list page size — see Output forms

Plus the connection flags and the output forms.

Examples

Every record in the namespace:

$ graphenectl get all
REF                    PHASE  OWNER        AGE    LABELS
agent/vm-e2e           ready  run/run-e2e  3m12s  role=e2e

pipeline/perf-nightly  ready               4d1h

Kinds are separated by a blank line: one long column becomes groups.

all is the installation's records; the dictionary of kinds is its own listing — get kind, or kinds.

One kind, filtered and watched:

$ graphenectl get docker-volume --owner stand/perf-nightly -w
REF                        PHASE  OWNER               LABELS
docker-volume/cache-v1     ready  stand/perf-nightly
docker-volume/cache-v1     ready  stand/perf-nightly  deleted

Runs by status:

$ graphenectl get run -p terminated
RUN         PIPELINE      STATUS      STARTED    TOOK   LABELS
watch-demo  perf-nightly  terminated  2h14m ago  1m48s  team=perf
val-c       perf-nightly  terminated  1d3h ago   42s

One record — the header fields, then the spec and the state as outlines. Empty fields and empty blocks are left out. A record's spec is whatever its kind's author made it (a docker container's is docker's whole Config), so the default view shows the first three levels and the first six items of a list, folds the rest into a one-line summary and says so; -o yaml is the whole record.

$ graphenectl get pipeline perf-nightly
ref:    pipeline/perf-nightly
phase:  ready
age:    4d1h
state:
  concurrency: queue
  digest: sha256:abc82597…
  image: localhost:7233/default/perf-nightly:4f925b8c6e5fff45
  manifest:
    activities: [… 14 items]
    params: {… 3 fields}
… long parts are folded; -o yaml shows the whole record

One run — what its listing row knows, spelled out:

$ graphenectl get run watch-demo
run:      watch-demo
pipeline: perf-nightly
status:   terminated
started:  2026-08-21 11:23:41
took:     1m48s
image:    localhost:7233/default/perf-nightly:4f925b8c6e5fff45
trigger:  manual
labels:   team=perf

Nothing, and no such thing

A listing shows live records. An empty answer says so, names the filters that narrowed it (on stderr — stdout stays clean), and exits 0:

$ graphenectl get docker-volume -p ready
No live docker-volume records match phase ready.

A kind the installation does not have is a mistake, not an empty set — the command fails and offers the nearest kinds:

$ graphenectl get agnt
graphenectl: unknown kind "agnt" — did you mean agent? (`graphenectl kinds` lists them)

The same line is drawn everywhere: get, delete, events, logs, run status, run result and run cancel answer no record <ref> (or no run <id>) with a non-zero exit for a target that does not exist, and never an empty success.

A record that finished its life is not in the listing, but can still be read by name — its last spec and state, with phase: deleted:

$ graphenectl get agent vm-e2e
ref:    agent/vm-e2e
phase:  deleted
...

One vocabulary of phases

A run's status and a record's phase are one field of one dictionary, lowercase: records go creating → ready → deleting → deleted (or failed), a run goes running → completed | failed | canceled | terminated | timed-out. The same words filter (-p failed), color and compare everywhere — Temporal's own spellings never reach the door.

A run keeps what it collected. get run <id> of a run that did not complete shows its state as the error and the partial result:

$ graphenectl get run nightly-0917 -o yaml
resource:
  ref: run/nightly-0917
  phase: failed
  state:
    error: "infrastructure suite failed on db-1: 1 failed, 2 passed"
    result:
      db:
        infra: {passed: false, diskMBps: 14.7}