Ultrareview CLI reference

Every command, flag and exit code the ultrareview CLI answers, generated from the command tree itself.

CLI reference

Generated from the command tree, so it cannot describe a flag that does not exist. For how to use these together, read Ultrareview from the terminal.

Commands

CommandWhat it does
ultrareview config explainShow every setting and which layer supplied it.
ultrareview config initWrite a starter .ultrareview, refusing to replace one that is already there.
ultrareview config schemaPrint the JSON Schema for that file, for an editor to complete against.
ultrareview config validate [<path>]Check a repository's .ultrareview file the way the review would read it. No network, no login.
ultrareview doctorCheck settings, reachability, scopes and expiry, all at once. Works signed out and offline.
ultrareview notes listThe notes your team has written, and what each one applies to.
ultrareview notes show <heading>One note, in full.
ultrareview pr findingsWhat a review found, and whether it should stop a build.
ultrareview pr historyEvery review of one pull request, newest first.
ultrareview pr reviewPosts a review comment your team will see, and once it starts nothing can stop it. The work runs on the server.
ultrareview pr statusWhat became of one review request.
ultrareview reviewReview the branch you are on, here, before there is a pull request. Uses your team's settings.
ultrareview reviews listRecent reviews across the organization.

Flags

FlagTakes a valueWhat it means
--allow-incompletenoReport a review that covered less than it should have, instead of failing on it.
--allow-stalenoRead a review of a different commit anyway.
--baseyesWhat to compare against. Defaults to the branch your remote calls its default.
--base-urlyesThe service host. Only a self-hosted installation needs this.
--fail-onyesLowest severity that fails a build, or never.
--forcenoReview the commit again even if it already was; replace an existing file (config init).
--formatyesHow to print: text, json or sarif.
--helpnoPrint help for the command and stop.
--hook-exitnoAnswer with the codes an agent hook speaks: 2 blocks, 1 reports without blocking, 0 proceeds. Without it a hook reads findings as a non-blocking error and a usage error as a block, which is the inversion in both directions.
--jsonnoShorthand for --format json.
--limityesHow many entries a list shows.
--modelyesReview on a different model, this once. Nothing is stored.
--no-contextnoReview without your team's settings, deliberately.
--pollyesHow often to ask again while waiting.
--pryesThe pull request number.
--profileyesThe review lens, by profile name.
--repoyesThe repository, as GitHub spells it: owner/name.
--runyesThe review run to read.
--stateyesShow only runs in one state.
--timeoutyesHow long to keep waiting before giving up.
--tokenyesThe API token. Prefer ULTRAREVIEW_TOKEN; a command line lands in shell history.
--versionnoPrint the version and stop.
--waitnoKeep watching until the review ends.

Exit codes

Every command answers with one of these, so a build script can act without parsing output.

CodeMeaning
0Nothing worth stopping for.
1Findings at or above your threshold.
2The command was used wrongly.
3The review covered less than it should have.
4Not authenticated: no token, expired, or revoked.
5The service or the network refused. Transient by nature, so this is the one worth retrying.
6This machine has no model access. A local condition, fixed locally, so retrying changes nothing.
7Authenticated, but not permitted. Ask an administrator; retrying changes nothing.
130Interrupted by a signal, following the shell convention of 128 plus the signal number.