CLI

Depot CLI: GitHub Actions commands

Use depot jobs to inspect GitHub Actions jobs running on Depot. Install the Depot CLI and authenticate.

Commands accept --org <org-id> and --token <token>. The organization defaults to your current organization. --output auto is the default: a terminal gets a table, and piped or redirected output gets JSON without a login prompt. Set --output table or --output json explicitly when needed.

depot jobs

List jobs newest first, or supply a GitHub job ID and exactly one --repo owner/name to inspect an individual job.

depot jobs --repo acme/api --output json
depot jobs <job-id> --repo acme/api

List queries default to the last 30 days and support a maximum 90-day window. To select a window, supply both --start and --end as RFC 3339 timestamps. The start is inclusive and the end is exclusive.

FlagDescription
--repoConnected repository in owner/name format; repeat to filter several repositories.
--querySearch job/workflow names or exact numeric job/run IDs.
--statusqueued, waiting, running, or finished; repeat to select multiple states.
--conclusionsuccess, failure, cancelled, skipped, or neutral; repeat to select multiple results.
--runner-labelRunner label; repeat to select multiple labels.
--start, --endCreation-time window for the job list.
--page-sizeResults per page, 1–250; defaults to 50.
--page-tokenReturned nextPageToken; keep the filters and time window unchanged.

The command returns one page at a time. Individual job inspection doesn't include step timings.

depot list jobs

An alternate entry point for depot jobs, with the same job-listing and inspection options:

depot list jobs --repo acme/api --output json

depot jobs metrics

Read CPU, memory, and out-of-memory evidence for one execution. A GitHub job ID and exactly one --repo are required. Completed jobs and snapshots of running jobs are supported.

depot jobs metrics <job-id> --repo acme/api --output json

JSON contains time-ordered samples, statistics, availability, and sample-cap metadata. CPU and memory utilization are fractions from 0 to 1 in JSON and percentages in tables. CPU is user-mode utilization averaged across cores; memory is used memory divided by total memory. Missing measurements remain absent in JSON and appear as n/a in tables.

Execution windows are limited to seven days and responses to 5,000 samples. Statistics cover all observed samples before downsampling, so the returned samples can omit intermediate peaks.

The JSON oom field reports out-of-memory evidence independently of CPU and memory sample availability. oom.detected is true when a kernel kill was observed and false when none were observed. When the lookup is unavailable, oom.detected is absent and oom.unavailableReason explains why. oom.events contains up to 100 observed kernel kills; oom.truncated indicates that more events were observed. Events can arrive after completion, so oom.detected: false doesn't rule out an out-of-memory failure.

depot jobs analytics

Show aggregate CPU and memory utilization, mean and p95 duration, and failure rate by repository, workflow, and job. Results are ordered by descending job count. The default window is 30 days; the maximum is 90 days. Analytics can lag completed jobs by up to 15 minutes and describe whole executions rather than individual steps.

depot jobs analytics --repo acme/api --job test --output json
FlagDescription
--repoRepository in owner/name format; repeat to select multiple repositories.
--workflowWorkflow display name; repeat to select multiple names.
--workflow-pathWorkflow file path; repeat to select multiple paths.
--jobJob key from the workflow file; repeat to select multiple jobs.
--runner-labelRunner label; repeat to select multiple labels.
--start, --endInclusive start and exclusive end in RFC 3339 format; supply both.
--page-sizeResults per page, 1–250; defaults to 50.
--page-tokenReturned nextPageToken; keep the filters and time window unchanged.

For test rankings across executions, use depot tests analytics.