CLI Command Reference
This page covers the Sigilix CLI — thesigilix binary you run in your terminal. For the @sigilix / /sigilix commands you type on a PR comment, see Slash Commands; those run on the hosted GitHub App, while the commands here run locally and reach for the same earned-context layer.
The Sigilix CLI is part of the private beta. Availability is per-account — join the private beta or contact support@sigilix.ai to get it enabled. The commands and flags below are representative; run
sigilix --help and sigilix <command> --help to see exactly what your account exposes.Command overview
| Command | What it does |
|---|---|
sigilix review | Runs the five-specialist ensemble on local changes, behind the believability gates |
sigilix chat | Opens Deep-Research Chat, grounded in your repo and review history |
sigilix config | Inspects and sets configuration — model selection, provider keys |
sigilix auth | Authenticates the CLI to your Sigilix account |
sigilix --help | Lists available commands and global flags |
sigilix review
Runs Metis, Argus, Iris, and Eunomia in parallel — unified by Harmonia — against your local changes, after the deterministic checks and behind the same believability pipeline a PR review uses. By default it diffs your working tree against the merge-base with the default branch.
Representative flags
Representative flags
--base <ref>— review the diff against this ref instead of the computed merge-base.--range <a..b>— review an explicit commit range.--path <glob>— scope the review to matching paths, on top of yoursigilix.jsonfilters.
sigilix review --help for the authoritative list on your account.Exit status
Exit status
The command exits non-zero when a blocking finding survives the gates, so you can wire it into a pre-push hook or a CI step as a fast first pass ahead of the authoritative hosted PR review.
What the output looks like
What the output looks like
A Harmonia summary followed by inline findings tagged by specialist (Metis / Argus / Iris / Eunomia), severity (Critical / Warning / Info), and proof tier (VERIFIED / GROUNDED / MODEL). See Sigilix CLI for a worked example.
sigilix chat
Opens Deep-Research Chat — plain-language Q&A grounded in the index, code graph, trust ledger, and review memory. Ask why a finding fired, what a change affects, or how a subsystem fits together; answers anchor to retrieved evidence rather than guesses.
What it can retrieve
What it can retrieve
Past findings, the specialist that raised them, the code they cited, their proof-tier receipts, and the rules you’ve taught — all from the earned-context layer. The chat reasons over these anchors instead of improvising.
Grounded by design
Grounded by design
When the earned-context layer doesn’t contain what a question needs, the chat says so rather than fabricating — the same believability bar as a review.
sigilix config
Inspects and updates CLI configuration. The most common use is selecting the model behind your terminal workflow — on paid tiers you can bring your own (Codex CLI, Claude Code, or your own SDK/keys). See Bring Your Own Models.
Model and provider names are surfaced by
sigilix config model for your account. Authentication follows each provider’s own flow — Codex CLI and Claude Code authenticate to OpenAI and Anthropic; “your own SDK / keys” uses the provider credentials you supply.sigilix auth
Authenticates the CLI to your Sigilix account so it can reach the earned-context layer for your repositories. Run this once after install.
Provider authentication (for the model you bring) is separate from Sigilix account authentication —
sigilix auth connects you to Sigilix; provider credentials are managed via sigilix config. See Bring Your Own Models.Read next
Sigilix CLI
Architecture and workflows behind
sigilix review.Deep-Research Chat
Architecture and workflows behind
sigilix chat.Bring Your Own Models
Configure Codex CLI, Claude Code, or your own keys via
sigilix config.Slash Commands
The
@sigilix / /sigilix commands on the hosted GitHub App.
