Ask better questions.
Keep control of access.

You can ask an AI assistant to check a room, compare settings, or help investigate a problem. The assistant runs commands through CE-Deploy and explains the results.

The gateway is the connection you switch on in CE-Deploy to allow those commands. It runs on your computer. The assistant needs access to that computer to use it.

You decide what is allowed

• Read only: the assistant can check information without changing devices.
• Read + write: the assistant can also make permitted changes.
• Write blocklist: commands and settings you have blocked stay blocked.
• Stop local gateway: turns off CLI access. CLI means command-line interface—the commands used in this course.

You do not need to share your tokens

CE-Deploy uses the Webex tokens and device passwords you have already saved in the app. Do not paste them into a chat or give the assistant credential files.

The assistant receives the information needed to answer your question. That may include private room or device details, so use an AI tool your organization allows.

Try it

Select AI & CLI Access at the top of the CE-Deploy window. Find Gateway, Organization access, Write blocklist, and Connect your tool. Leave write access off.

↑ All lessons

We use Codex and ChatGPT for the setup examples. You can use Claude Code, GitHub Copilot or another assistant with the same CE-Deploy commands. It needs permission to run commands on the same computer and computer account as CE-Deploy. A chat without that access cannot use the local gateway.

1. Turn on read-only access

1. Open AI & CLI Access → Organization access.
2. Choose your training organization, select Read only, then Save access.
3. Open Gateway and select Start local gateway. Keep CE-Deploy open.

Access lasts 30 minutes by default. If it expires while you read, start the gateway again before the next exercise. Refresh access checks the status; it does not add more time.

2. Ask your assistant to find the commands

If it cannot find the launcher, open AI & CLI Access → Connect your tool. Under Terminal, select Copy command. Check the installation path before giving it to your assistant: Windows shows an example path you must replace, and the macOS example assumes the app is in Applications. Use the launcher inside the same installed app you have open. You can also select Open CLI & AI guide there.

3. Add the supplied skill, if you want

A skill saves you from explaining the workflow in every chat. It is optional; the assistant can read the AI guide instead. The prompt below is a Codex example.

Skill setup differs between tools. If you use Claude Code, Copilot or another assistant, follow its instructions for adding a skill, or ask it to read CE-Deploy’s ai guide. The CLI commands and CE-Deploy permissions are the same either way.

4. Choose the organization

Check the answer: it must name the organization you intended. Changing the organization shown in the CE-Deploy window does not change the assistant’s choice.

On a narrow screen, swipe the image sideways to read it.

"/Applications/CE-Deploy.app/Contents/Resources/cli/ce-deploy" --help --json
ce-deploy() { "/Applications/CE-Deploy.app/Contents/Resources/cli/ce-deploy" "$@"; }

& "C:\path\to\CE-Deploy\resources\cli\ce-deploy.cmd" --help --json
$ceDeployCli = "C:\path\to\CE-Deploy\resources\cli\ce-deploy.cmd"
function ce-deploy { & $ceDeployCli @args }

ce-deploy organizations list --json
ce-deploy context select --organization ORG_REF --json
ce-deploy context get --context CONTEXT_REF --json

ce-deploy ai skill export --output .agents/skills/ce-deploy-access --json
ce-deploy ai skill check --file .agents/skills/ce-deploy-access/compatibility.json --json

If you get stuck

• Launcher not found: copy the path from Connect your tool.
• Organization missing: check that it is connected and that you saved Read only access.
• App unavailable: keep CE-Deploy open and check that the gateway is running.

OpenAI’s skill setup guide · Claude Code · Copilot CLI

↑ All lessons

CE-Deploy checks whether a command is allowed before it runs. The assistant cannot grant itself permission through the CLI.

Three reasons a command may be refused

• Access is off or has expired. Check the gateway in CE-Deploy.
• The organization is Read only. The assistant can check information, but cannot change a device.
• The action is blocked. The saved Write blocklist prevents it, even when Read + write is enabled.

Some commands also cannot be used as read-only checks. A command called “Get” or “List” is not automatically safe. If it is refused, the assistant should find an allowed way to read the information or explain what it could not check.

Review your blocked commands

Open AI & CLI Access → Write blocklist. Each entry shows an action name and its command or setting. Starter entries include factory reset, reboot and log deletion. Your saved list may differ.

For this lesson, look for the reboot entry. If it is absent, note that rather than assuming it is blocked. Do not run it or edit the list.

On a narrow screen, swipe the image sideways to read it.

A rule can block one command, a whole group of settings, or an action only when particular options are used. For example, deleting logs can be blocked while stopping logging without deletion remains allowed.

Read + write is permission to make allowed changes. CE-Deploy does not show a separate approval box for every job, so leave it off until you need it.

↑ All lessons

Example: people report quiet audio in Boardroom East before important meetings. Check its volume settings without changing them.

Replace Boardroom East and Example Corp with your own room and organization in the prompts below.

1. Find the right device

1. Keep CE-Deploy open and check that the gateway is running. Start it again if access has expired.
2. Ask the assistant to select your organization again if needed.
3. Give it the room name. Ask it to show the matching device before reading its settings.

Check the name and model in the reply. If the model is unavailable, confirm the Device ID instead. If the device is wrong, stop and correct the choice. If you already know the Device ID, give that to the assistant to help it find the right device.

2. Read the volume settings

Current volume is the volume reported now. Default volume is a saved setting. They are different values; a change to one does not prove the other changed.

3. Read the answer carefully

A useful answer names the room, says when it was checked, shows the returned values, and lists anything the assistant could not check. A failed read does not prove the device is offline.

On a narrow screen, swipe the image sideways to read it.

4. Save the report for next time

Keep a dated note with the organization, room, Device ID, returned values and failed checks. Store it somewhere your organization allows.

If this is your first check, say so. You can compare later checks with it, but do not call a faulty room’s settings “known good.” Ask the room’s owner which earlier report or working setup to use for comparison.

The assistant can spot a changed value. It cannot prove what people hear in the room. An attended audio test may still be needed.

ce-deploy devices list --context CONTEXT_REF --json

ce-deploy devices list --context CONTEXT_REF --device-id DEVICE_ID --json
ce-deploy xapi search --query "audio volume" --json
ce-deploy xapi schema --operation Status:Audio.Volume --json
ce-deploy xapi schema --operation Configuration:Audio.DefaultVolume --json

{
  "selectionRef": "SELECTION_REF",
  "requests": [
    {
      "type": "Status",
      "path": "Audio.Volume"
    },
    {
      "type": "Configuration",
      "path": "Audio.DefaultVolume"
    }
  ]
}

ce-deploy queries submit --context CONTEXT_REF --request-id UNIQUE_REQUEST_ID --file query.json --json
ce-deploy queries results QUERY_REF --wait 5 --json

ce-deploy queries results QUERY_REF --wait 5 --cursor CURSOR --json

↑ All lessons

We use Dots, ChatGPT’s assistant offering, for this example. With computer access, a dot can arrange checks on your CE-Deploy computer when you message it from elsewhere or ask for a schedule.

Other providers may call their features agents, assistants, remote control or scheduled tasks. For example, Claude Code’s Remote Control lets you continue a local session from another device. Remote access and scheduling are separate features; check which your tool supports. The same CE-Deploy CLI and access rules still apply.

First complete Lesson 4 and save its report. Give your assistant the same organization and Device ID so it checks the same room. The steps below show the Dots setup.

1. Connect the computer

1. On the CE-Deploy computer, open your dot’s profile in ChatGPT.
2. Under Computers → Your computer, select Allow access and review the confirmation.
3. Keep the computer awake and online, with ChatGPT and CE-Deploy open.
4. In CE-Deploy, save Read only for the organization and start the gateway.

This depends on the features and permissions available to your account. Connecting your computer to another ChatGPT tool does not automatically give your dot access.

2. Try one check from your phone

Check the reply against your saved report. Make sure it checked the same device and included the time.

3. Add a schedule, if useful

Review the saved task in Scheduled and check the first run. Give the task the room details and report location; do not assume a new task knows everything from another chat.

In AI & CLI Access → Gateway, choose a Gateway duration and select Save duration. The choices are 30 minutes, 1 hour, 4 hours, 8 hours, or Until CE-Deploy quits. If stopped, select Start local gateway. If already running, Restart with saved duration applies the new choice; ask the assistant to reconnect afterward. Refresh does not renew access.

Leave the organization on Read only while you are away. CE-Deploy provides no scheduled-write mode, but a local assistant could still use shared Read + write access if you leave it enabled.

When monitoring ends

Cancel the schedule in Scheduled and stop the gateway. If the dot no longer needs your computer, revoke its computer access separately.

OpenAI guides: Connect your computer · Recurring tasks

↑ All lessons

Use a lab device and an agreed change window. You can finish this lesson by preparing a plan without changing a device. You do not need to complete the Dots lesson first.

Example: the person responsible for the lab asks you to set its default volume to 50. This is practice, not a proposed fix for Boardroom East.

1. Ask for a plan first

Check the device and values. If the setting is already 50, there is nothing to change. If the original value cannot be read, stop and agree on a recovery plan before continuing.

2. Allow the reviewed change

When you are ready, open AI & CLI Access → Organization access, choose Read + write for that organization, and select Save access. Review the Write blocklist and check that the gateway is running. Ask the assistant to select the organization and device again after this permission change, then carry out the agreed plan.

If the app says write execution is disabled, stop here and keep the exercise as a plan. CE-Deploy does not ask for approval again for each job. Permission settings are the real control; the instruction to wait is part of your conversation with the assistant.

3. Check what happened

Ask the assistant to collect all results, then read the default volume again on the same device. Open History at the top of CE-Deploy. Find the entry for your change, labeled AI / CLI, and select its row to view the details. Match its time and room to your change.

• Acknowledged: check the setting again before calling the change successful.
• Error: inspect the result and do not assume nothing changed.
• Unknown: the change may have reached the device. Investigate; do not automatically run it again.

Stopping a job prevents remaining work where possible. It cannot undo a change already sent. Restoring the old value is a separate change to review and check.

4. Turn write access off

In Organization access, choose Read only again and select Save access. To remove access entirely, choose Disabled and save it. In Gateway, select Stop local gateway when finished. Record the old value, requested change, returned results and readback in your report.

{
  "selectionRef": "SELECTION_REF",
  "operations": [
    {
      "type": "Configuration",
      "path": "Audio.DefaultVolume",
      "value": 50
    }
  ]
}

ce-deploy writes submit --context CONTEXT_REF --request-id UNIQUE_REQUEST_ID --file change.json --json
ce-deploy writes status JOB_REF --context CONTEXT_REF --json
ce-deploy writes results JOB_REF --context CONTEXT_REF --wait 5 --json

{
  "selectionRef": "SELECTION_REF",
  "requests": [
    {
      "type": "Configuration",
      "path": "Audio.DefaultVolume"
    }
  ]
}

ce-deploy writes cancel JOB_REF --context CONTEXT_REF --json

↑ All lessons