Xyte CLI · for Claude, Codex, and Copilot

Ask your fleet questions. Get real answers.

Xyte CLI connects the AI assistant you already use to your Xyte fleet. Set it up once, then copy a prompt below to get a daily fleet report, a firmware report, or a morning check that recovers offline rooms before anyone walks in.

Set up

Three steps. No terminal skills needed.

You'll need a Xyte account and an AI assistant that can run commands on your computer, such as Claude, Codex, or GitHub Copilot in Agent mode.

1

Create an API key

In Xyte, go to Settings → API Keys and create a key. Save it in a text file on your computer, outside any project folder.

2

Paste the setup prompt

Copy the prompt below into your assistant. It installs Xyte CLI with its skills and checks that everything works.

3

Ask for what you need

When setup is done, pick a prompt from the library below, or just ask a question about your fleet in your own words.

Setup prompt

Paste this into your assistant once. It takes a few minutes.

Run once
Install Xyte CLI (@xyteai/cli) with all of its skills from https://github.com/xyte-io/xyte-cli. 1. Check my computer has what it needs (Node.js 22 or later) and install anything that's missing. 2. Make xyte-cli work from any folder, not just this one. 3. Connect my Xyte API key. It's saved in a file at [path to your API key file]. Don't ask me to paste the key into chat, don't print it, and don't save it inside a project folder. 4. Confirm the connection works and tell me my tenant name. 5. Run a quick read-only check and tell me how many devices and spaces you can see. Never make up IDs, numbers, or results. If something fails, tell me what went wrong in plain language.

You'll get

  • Xyte CLI installed and working from any folder
  • Your API key connected securely
  • A quick count of your devices and spaces

Tip

  • Replace the [highlighted text] before sending
  • Restart your assistant afterward if it asks you to, so it picks up the new skills
Keep your key private.Your API key gives access to your fleet. Point your assistant to the file instead of pasting the key into a chat, email, screenshot, or recording.
Prompt library

Copy, paste, get the answer.

Each prompt works as written. Fill in the highlighted parts, or leave them out to cover your whole fleet. Change anything you like: your assistant understands plain language.

Daily fleet report

A one-page status of your whole fleet for a morning check-in, a team handoff, or a customer update.

Read-only
Use xyte-cli to create today's fleet report for [customer name, or "all spaces"]. Include: - Total devices, and how many are online vs. offline (with the % offline) - Number of spaces - Active and closed incidents - Open tickets - The 10 rooms with the most problems, and what's wrong in each - Anything that changed significantly since yesterday Save the raw data as JSON, then turn it into a clean one-page PDF I can forward to [my team / the customer]. Put the rooms that need attention first at the top. Only use real data from xyte-cli. Don't guess or fill in missing numbers.

You'll get

  • A PDF you can forward as-is
  • A JSON file with the source data
  • A short summary in chat

Make it yours

  • "…and email it to me at [address]"
  • "…and make this run every weekday at 8 AM" (see Automations)

Firmware vulnerability report

See which firmware versions are running across your fleet, and where unknown or out-of-date firmware puts rooms at risk.

Read-only
Use xyte-cli to build a firmware vulnerability report for [customer name, or "all spaces"]. For every device, pull the manufacturer, model, room, and current firmware version. Then: - Group devices by model and show which firmware versions are in use, and how many devices run each one - Flag devices where the firmware version is unknown or missing - Flag devices running an older version than other devices of the same model - Rank the highest-risk rooms first (most devices with unknown or older firmware) - End with a short list of recommended next steps Add a note at the top that this report shows firmware exposure and update readiness. It is not a confirmed CVE check. Save the source data as JSON and create a PDF.

You'll get

  • A highest-risk rooms table
  • Firmware version spread by model
  • A PDF plus the JSON inventory

Good to know

  • Xyte shows what firmware is running. To match versions to specific published vulnerabilities, check each manufacturer's security notices.

Executive health summary

A short, non-technical summary for leadership or a quarterly business review.

Read-only
Use xyte-cli to write an executive summary of the health of [customer name, or "our AV environment"]. Keep it to one page and plain language. Cover: overall uptime and health, the locations with the most issues, the biggest risks right now, how things have trended over the last [30] days, and 3 recommended next actions. Make it a PDF.

You'll get

  • A one-page PDF in plain language

Make it yours

  • "…and match our brand colors"
Automations

Fix offline rooms before the first meeting.

Schedule a prompt to run on its own every morning. Test it first with a dry run, approve what it will do, then put it on a schedule.

1

Do a dry run

Paste the prompt and ask your assistant to show you what it would do, without sending any commands.

2

Approve the plan

Check the rooms, devices, and commands it lists. Ask for changes until you're happy with them.

3

Schedule it

Ask your assistant to run it on a schedule, or add it under Scheduled tasks in Claude or Codex.

Morning room recovery

Checks every room, tries a safe reboot on devices that are offline or in an error state, confirms whether they recovered, and emails you a report.

Sends device commands
Every day at [8:00 AM], use xyte-cli to check every room in [customer name, or "all spaces"]: 1. Find every device that is offline or has an active error incident. 2. For each one, check which recovery commands it supports: reboot, restart, power cycle, or power on. Only use those. Never change settings, firmware, or configuration. 3. Send one recovery command per device. 4. Wait [5] minutes, then check each device again. 5. Mark a device as recovered (green check) ONLY if it is back online or its incident has cleared. Otherwise mark it "not recovered." Then email a report to [you@company.com] every time this runs, with: - Rooms checked, rooms affected, and devices affected - Every command sent, and whether the device accepted it - Which devices recovered and which didn't - Devices that have no recovery command (these need a person), listed separately - Attach the report as a PDF Before the first scheduled run, do a dry run: show me exactly which devices you would send commands to, and which commands, and wait for my OK.

You'll get

  • A room-by-room PDF with recovered / not recovered status
  • A short list of devices that need a person
  • An email every morning

Make it yours

  • "…only on weekdays"
  • "…skip rooms that are booked in the next 30 minutes"
  • "…if a device fails 3 days in a row, open a ticket"

Scheduled daily fleet report

The daily fleet report, delivered to your inbox every morning. Nothing on your devices changes.

Scheduled · read-only
Every weekday at [7:30 AM], use xyte-cli to create a fleet report for [customer name, or "all spaces"]: devices online vs. offline, active incidents, open tickets, the 10 rooms that need attention first, and what changed since the last report. Make it a one-page PDF and email it to [you@company.com]. Only use real data from xyte-cli. Don't change anything on any device.
You stay in control.Scheduled prompts use the same API key and permissions as you, and every command shows up in Xyte's audit trail. Start with a small group of rooms, read the first few reports, then widen the scope. xyte-cli can't send email itself, so the email step works only if your assistant can send mail. If it can't, ask it to save the PDF to a folder instead.
Tips

Get better answers

A few habits that make every prompt work better.

Start with "Use xyte-cli"

This tells your assistant to get live data from your fleet instead of answering from general knowledge.

Say who it's for

"For my team" gets detail. "For the customer" or "for leadership" gets a cleaner, shorter report.

Ask for a file

Add "make it a PDF" or "save it as a spreadsheet" to get something you can forward right away.

Look before you change

Ask questions and run reports first. When you want changes, ask for a dry run so you can approve the plan.

Ask for real data only

"Don't guess or fill in numbers" keeps reports accurate. AI can still make mistakes, so check anything important.

Keep your skills current

Every so often, ask: "Update your Xyte CLI skills from github.com/xyte-io/xyte-cli."

Watch

See Xyte CLI in action

A quick walkthrough of setup and your first fleet report.

Help

Common questions

Most problems can be fixed by telling your assistant what happened.

My assistant says "xyte-cli: command not found"

Xyte CLI is installed but your computer can't find it from this folder. Ask your assistant: "Make xyte-cli work from any folder, then check it in a new terminal."

It says I need Node.js 22 or later

Xyte CLI runs on Node.js. Ask your assistant: "Install or update Node.js to version 22 or later, then continue the Xyte CLI setup." On a work computer, you may need IT to approve the install.

The report is missing numbers or looks wrong

Ask: "Which of these numbers came directly from xyte-cli, and which didn't?" Then ask it to re-run using only xyte-cli data. If some devices don't report a value (like power or firmware), ask for them in a separate list.

Which AI assistants work?

Any assistant that can run commands on your computer: Claude, Codex, and GitHub Copilot in Agent mode. Chat-only assistants can't run Xyte CLI.

Is it safe to let an AI assistant send commands to devices?

Your assistant can only do what your API key allows, and every action is logged in Xyte's audit trail. Reports and questions are read-only. For anything that changes a device, use a dry run first and approve the plan. Keep scheduled automations to safe recovery commands like reboot or power on.

I prefer the terminal. What are the commands?

Requires Node.js 22+.

# check your environment and follow its recommendation
npx -y @xyteai/cli@latest doctor environment --format json

# connect your API key and confirm
xyte-cli setup run
xyte-cli setup status --field tenantId

# set up agent skills in this project
xyte-cli init --scope project --agents all --no-setup

For CI and headless jobs, pin a version instead of @latest. See the CI and agents guide and the commands reference.

Your first fleet report is five minutes away.

Create an API key, paste the setup prompt, and ask your first question.