@qoren/cli

Drive your Qoren account, environments and agents from a terminal.

npm install -g @qoren/cli
qoren login
qoren            # the whole account on one screen
qoren env ls     # or one answer at a time

This is the short version. The full user guide, covering the sign-in flow, every command, scripting, configuration and troubleshooting, is at https://qoren.sh/docs/cli.

Interactive mode

Run qoren with nothing after it at a terminal and it opens full screen: environments, agents, jobs, a chat pane and your account, all on the keyboard. qoren tui is the explicit form. Everything the commands can do it can do, through the same credential and the same plan limits.

tab / 1-5     move between sections     enter   open what is selected
up down j k   move within a list        r       refresh
?             every key, in context     q       quit

Two things it does that a command cannot. A reply streams in as the agent writes it, naming the tool it is running, instead of a spinner and then a wall of text. And when a region cannot run the size you asked for, it offers you the region that can, rather than printing a flag to re-run with.

Piping or a non-interactive shell is unaffected: qoren with no terminal still prints the help text, and --json is refused rather than half-honoured.

Commands

qoren login [--token] [--name <label>]   sign this computer in
qoren logout [--all]
qoren whoami

qoren env ls | get <id> | vitals <id>
qoren env create --name <n> [--size <s>] [--region <r>] [--auto-region]
qoren env rm <id> | rename <id> <name> | resize <id> --size <s>
qoren env collaboration <id> on|off

qoren agent ls [--env <id>] | get <id>
qoren agent create --env <id> --template <slug> --name <n> [--model <m>] [--runtime <r>]
qoren agent rm <id> | rename <id> <name>
qoren agent message <id> "<text>" [--resume <sessionId>]
qoren agent exec <id> "<command>"
qoren agent logs <id> [--limit <n>] | status <id> | telemetry <id> [--days <n>]
qoren agent doctor <id> [--no-repair] [--no-wait] | doctor <id> --history
qoren agent chatgpt-login <id> [--no-wait] | chatgpt-logout <id>

qoren account usage | spending [--days <n>] | options | templates

qoren jobs ls | get <id> | watch <id> | cancel <id>

qoren api <METHOD> <path> [--data <json|@file>] [--query k=v]

qoren tui                                open interactive mode

Run qoren <command> --help for full options.

Regions

Leave --region off and the platform picks one that can run the size you asked for. Pass --region and it is pinned: a region you named is never swapped for another, so if it cannot run that size the create is refused rather than quietly moved. The refusal names the closest region that does have it, and --auto-region on a re-run is your permission to use it. It also lists the sizes your region does have, in case staying put suits you better.

qoren env resize is refused the same way when the region an environment already lives in cannot run the target size. There is no --auto-region there: an environment cannot change region, so the refusal lists the sizes that region can run and one of those is the way forward.

Scripting

Every command takes --json. In that mode stdout carries only the payload; progress and warnings go to stderr, so piping is safe.

qoren env ls --json | jq -r '.[].name'

Exit codes:

Code	Meaning
0	success
1	the request failed
2	a mistake in the command
3	sign in again (missing, expired or revoked credential)
4	the plan does not allow this, including a plan without API access

The CLI is API access, which the Ultimate, Business and Enterprise plans include. On another plan every command exits 4 and names the plans that include it; see https://qoren.sh/pricing. The web console works on every plan.

CI

Create a token under Settings, CLI tokens, and set it as QOREN_TOKEN. It overrides any stored login, so a pipeline can never pick up a developer's profile.

export QOREN_TOKEN=qrn_...
qoren agent create --env "$ENV_ID" --template deploy-bot --name "Deploy bot" --json

Configuration

Variable	Effect
QOREN_TOKEN	token to use; the config file is not consulted
QOREN_API_URL	server to talk to (default https://qoren.sh)
QOREN_PROFILE	which stored login to use
NO_COLOR	disable colour

Credentials live in ~/.config/qoren/config.json (or $XDG_CONFIG_HOME/qoren), written 0600. It holds bearer tokens: treat it as a password. Named profiles let you hold several logins; switch with --profile.

Jobs

Creating an environment or deploying an agent is asynchronous. The CLI follows the job and prints each step as it happens. Add --no-wait to get the job id immediately and follow it later with qoren jobs watch <id>.

Anything not yet a command

qoren api reaches any control-plane endpoint with the same credential and the same rules, so a missing command is an inconvenience rather than a blocker.

qoren api GET fleetSummary
qoren api POST machines --data @environment.json

Development

npm install
npm run check   # eslint + tsc
npm test
npm run build && node dist/index.js --help

Interactive mode is Ink (React on a terminal). Its logic lives in src/tui/model.ts as pure functions so it can be tested without a screen; src/tui/model.test.ts is where its behaviour is pinned down. Ink and React are loaded dynamically, so a scripted qoren env ls --json never pays for them.

Commands live in src/commands/ and call @qoren/sdk; interactive mode lives in src/tui/ and calls the same SDK. Neither ever builds a URL itself. That is the rule that keeps the CLI and the SDK in step: an endpoint is described once, in the SDK, and every client reads it there.