PlatformCLIGovernanceQLoop

Top 15 quox CLI commands, with examples

Adam Cowles17 August 20264 min read

A dark terminal prompt with a cyan quox command glowing mid-keystroke

Your terminal is the control plane for a governed multi-agent platform. Not a thin wrapper over an API: the whole of QuoxCORE (agents, fleet, autonomous loops, memory, evidence, roadmap) is drivable from one quox binary.

The part people do a double-take at: when a governed agent hits an action it is not allowed to take on its own, it does not stop and it does not sneak around the rule. It hands you the exact command to release the gate, and you paste it. Governance you can run from a prompt, with a witnessed evidence trail behind every line.

Here are the fifteen we reach for most.

1. quox login

Authenticate once, everything else inherits the session.

bash

quox login

Profiles and orgs are first class: quox --org acme login and quox -p staging login keep separate contexts side by side.

2. quox status

The whole platform's health in one screen: services, queues, workers.

bash

quox status

When something feels off, quox doctor runs the deeper diagnostics.

3. quox chat

Talk to CommanderQ or route straight to a named agent, streaming by default.

bash

quox chat "which workflows failed in the last hour and why?" --agent cipher

No dashboard needed to ask the estate a question.

4. quox fleet exec

Run a governed tool against a managed host. Every call is audited, no raw SSH.

bash

quox fleet exec --tool health-check --input '{"host":"web-prod-01"}'

quox fleet status and quox fleet agents show what is under management first.

5. quox objective create --execution-mode loop

Stand up an autonomous agentic loop with an explicit capability scope.

bash

quox objective create "triage new gripes and draft fixes" \
  --execution-mode loop --loop-backend claude \
  --capability-scope repobrain_query,gripe_read

A DAG plan is the default. --execution-mode loop runs a QLoop backend instead, and because a loop is not a reviewable plan, it always asks for one human approval before it starts.

6. quox objective checkpoints + approve-checkpoint

This is the one. A running agent reaches a gated step. It cannot self-approve, so it surfaces the exact command for a human to run:

bash

quox objective checkpoints
quox objective approve-checkpoint <objectiveId> <checkpointId>

The gate releases, the loop continues, and the approval is witnessed on the evidence chain. Deny with deny-checkpoint, or stop a running loop outright with quox objective halt <id> (same signed kill as a denial).

7. quox inbox respond

The same human-in-the-loop pattern for one-off tool runs and requests.

bash

quox inbox list --status pending
quox inbox respond <itemId> --action approve --comment "verified with owner"

The approval queue lives in your terminal, not a tab you forgot to open.

8. quox memory search

Durable, typed memory your agents actually read before re-deriving anything.

bash

quox memory search --query "postgres connection tuning"
quox memory create --content "..." --type fact

9. quox brain

Semantic search across vaults and notes when you want the knowledge, not a keyword match.

bash

quox brain search "how does our billing retry logic work?"

10. quox done check

DoneEngine answers the only question that matters before a release: can this ship, and what is the evidence? Advisory, never blocks.

bash

quox done check <subject>

quox done board is the estate scoreboard across every surface.

11. quox volt

Inspect and verify a VOLT evidence bundle. Proof that a run happened the way the logs say.

bash

quox volt verify <bundleId>

Paired with quox governance for policies and the WARD hash chain, this is the audit trail, self-serve.

12. quox trace

Replay and visualise an agent run end to end: what it saw, what it decided, what it called.

bash

quox trace list
quox trace show <traceId>

13. quox plan focus

QuoxPlan's focus lens: what matters right now (P0, blocked, decision needed, in flight, stale).

bash

quox plan focus

quox plan show <streamId> opens the full stream, and notes are append-only so the history is intact.

14. quox skills

Query, run, and promote governed skills (QSKL Crystals) from the shared library.

bash

quox skills list
quox skills run <crystalId>

15. quox qlarity read

Ask a live page a question through the governed browser surface, session isolated and receipted.

bash

quox qlarity read https://app.example.com --inspector nova

The safe way to let an agent look at a running page, with arbitrary evaluation always gated behind a human approval.

Try it yourself. Every command above runs in the QuoxCORE web terminal, no local install required. The approval flow (commands 6 and 7) is the one to watch: start a loop, let it reach a gate, and see it print the exact approve-checkpoint line for you to run.

Full command reference is in the docs, and the QuoxCLI product page has the quick start.

Drive the whole platform from one prompt

QuoxCLI ships with QuoxCORE. Every command is governed, witnessed, and yours to script.

See QuoxCLI Read the docs