
Use **AI-Pallet QA and Routing** when you want a pallet-focused AI workflow for inspection, routing, and quality decisions.

This option is best when the workflow is already centered around pallet handling and you want to begin from a purpose-built AI template instead of a blank schema.

## Choose AI-Pallet QA and Routing

1. Open a channel.
2. Click the **Data Sources** tab.
3. Click **Add Data Source**.
4. Choose **AI-Pallet QA and Routing**.

![Add Data Source dialog showing AI-Pallet QA and Routing as one of the available types.](/docs/data-sources/ai-pallet-qa-and-routing/01-type-selection.png)

## Start from the pallet template

1. Review the available pallet templates.
2. Select the one that matches the workflow you want.
3. Enter the data source name.
4. Click **Create data source**.

This path gives you a faster setup for pallet QA and routing work, so your team can begin from a workflow that already matches the device purpose.

## Continue in the data source workspace

After creation, open the data source and review the structure that came with the selected template. From there, you can continue into the routing or QA pages that match your process.

## Continue with guided setup

After the Pallet QA data source exists, you can use **Home AI Assistant** to guide setup and model improvement.

Ask:

- set up vision device
- what is next for this pallet vision data source?
- help me improve this pallet vision model

The assistant checks whether the device is paired and mounted, then guides you to the next useful tab.

For Pallet Vision, snapshots are usually uploaded automatically by the device. Because of that, the guided workflow focuses on:

1. pairing and mounting status
2. checking **Live View** when needed
3. waiting for or reviewing validation snapshots
4. copying useful examples into a dataset
5. annotating and versioning the dataset
6. training a model
7. pushing the model to the device
8. reviewing fresh validation snapshots and inspection results

Use the assistant again after each milestone if you want it to suggest the next step.

## Read next

- [Create a Data Source](/docs/data-sources/create-a-data-source) - Return to the main create flow.
- [Home AI Assistant](/docs/ai-assistance/home-ai-assistant) - Use guided setup for an existing Pallet Vision data source.
- [Image Review](/docs/data-sources/image-review) - Review QA session images with overlays, inspection details, and marks.
- [Video Review](/docs/data-sources/video-review) - Review recorded validation clips and add marks at the right moment.
- [Routing](/docs/ai-vision/routing) - Continue with routing-related work after setup.
- [Data Sources Tabs](/docs/data-sources/channel-data-sources) - Open the created data source from the channel table.


Create an AI-Pallet QA and Routing data source when you want a pallet-focused AI workflow with a guided starting template.

AI-Pallet QA and Routing


Connect a site to **Claude** and you can ask about your operation the way you would ask a
supervisor. Claude looks the answer up in your live Zira data — you do not copy readings into a chat
window or paste answers back out.

> **In beta.** This is new and we are still refining it. It works, and we would like to hear how it
> goes for you.

## Before you start

- You need to be signed in to Zira, and you need access to the site you want to connect.
- You need Claude — the desktop app, Claude Code, or another assistant that accepts a connection
  address.
- **One connection covers one site.** If you look after several sites, connect each one you want
  Claude to reach.

## Connect a site

1. Click your **profile picture** in the top right and choose **Connect to Claude**.
2. Under **Which site?**, search for and pick the site.
3. Under **What can Claude do?**, choose **Look only** or **Look and act**.
4. Click **Connect to Claude**.
5. Pick your assistant from the tabs — **Claude Code**, **Claude Desktop**, or **Something else** —
   then click **Copy** and paste what you copied where the page tells you.

![The Connect to Claude panel with a site chosen and Look and act selected.](/docs/ai/connect-to-claude/01-connect-panel.png)

That is the whole setup. Claude Code picks the connection up in your next session; Claude Desktop
reads it when it starts, so restart it.

## Choose what Claude may do

| Choice | What it means |
| --- | --- |
| **Look only** | Claude can read your data and nothing else. It cannot create, change, or post anything. |
| **Look and act** | Claude can create and update tasks, post readings, and post to channels — and asks you before each one. |

**Look only** is a good way to start. You can come back and connect again with **Look and act** once
you are comfortable.

## Copy the connection line before you close the page

The connection line is shown **once**. Copy it straight away — if you lose it, come back and connect
again.

Treat it like a password. Anyone who has it can reach that site. If you stop using a connection, or
you think the line has leaked, remove it from that site's **Applications** tab.

## Ask about your operation

Once connected, ask the kind of question you would ask on the floor:

- *"Why did the trim saw stall last night? Check its readings around 22:00."*
- *"What was yesterday's output on line 3?"*
- *"Show me the open tasks for the Rehovot plant."*

Claude finds the right data source by name and looks the readings up itself. You do not need to know
which metric or device to point it at.

## Let Claude do the work

With **Look and act**, Claude can open and update tasks, record readings, set a device's status, and
post an update to a channel.

Claude asks you before each one. Treat that as a real confirmation — **creating a task notifies the
person it is assigned to, and posting to a channel is visible to everyone following it.**

## Tips

- **Ask for names, not IDs.** "The bandsaw in Plant 2" is enough — Claude searches for it.
- **Keep time ranges narrow.** "Last night between 21:00 and 23:00" gets a fast, precise answer;
  "this year" returns more than Claude can read at once.
- **It cannot see everything.** A connection covers data sources, readings, tasks and files.
  Dashboards, forms and vision models are not reachable this way.

## Let Claude read the guides too

There is a second, separate connection that gives Claude the Zira help center — the product guides,
the API reference, and the release notes. It is public and read-only, so it needs no setup beyond
the address. Teams that add both get answers grounded in how Zira actually works.

## Read next

- [Connect Zira to Claude — step by step](/developers/how-to-guides/connect-zira-to-claude)
- [Connect the documentation](/developers/how-to-guides/connect-the-documentation-mcp-server)
- [Home AI Assistant](/docs/ai-assistance/home-ai-assistant)
- [Tasks](/docs/tasks/channel-tasks)


Connect a site to Claude and ask about your operation in plain language — Claude reads your live Zira data, and can open tasks and post updates when you let it.

Connect Zira to Claude


Use **AI-Vision** when you want a ready-made setup for counting, measuring, monitoring, or inspecting objects with a Zira AI-Vision device.

This type is helpful when you want to move quickly from setup to live device pairing without building the data source from scratch.

## Start from the Data Sources tab

Open the channel where you want to add the device, then open **Data Sources**.

You will see the list of existing data sources for that channel.

![Data Sources tab showing the list of existing data sources in the channel.](/docs/data-sources/channel-data-sources/01-data-sources-list.png)

Click **Add Data Source** in the top-right area to begin.

![Add Data Source button in the top-right area of the Data Sources tab.](/docs/data-sources/channel-data-sources/02-add-data-source-button.png)

## Choose AI-Vision in the create flow

1. Open a channel.
2. Click the **Data Sources** tab.
3. Click **Add Data Source**.
4. Choose **AI-Vision**.

![Add Data Source dialog showing AI-Vision as one of the available types.](/docs/data-sources/ai-vision-data-source/01-type-selection.png)

The AI-Vision card is the best choice when you want a fast setup based on a built-in vision template.

## Pick a template and name the data source

1. Review the available templates.
2. Select the template that matches your workflow.
3. Enter the data source name.
4. Click **Create data source**.

![AI-Vision setup showing the data source name field and available templates.](/docs/data-sources/ai-vision-data-source/02-template-selection.png)

Pick the template that feels closest to the workflow you want to track. You can still review and adjust the setup after the data source opens.

## Review the built-in fields

After the data source opens, Zira shows the fields that came with the selected AI-Vision template.

These fields usually give you a strong starting point for things like:

- counts or units
- status values
- shift or event tracking
- measurements such as length or width
- product or people-related entries

![AI-Vision data source editor showing the built-in fields and live form preview.](/docs/data-sources/ai-vision-data-source/03-fields-and-preview.png)

This screen is helpful because it lets you see both sides of the setup at once:

- the field structure on the left
- the live result on the right

Before pairing the device, take a moment to confirm that the fields match what you want operators or workflows to capture.

## Pair a Zira AI-Vision device

When you are ready to connect the device:

1. Open the data source row actions menu.
2. Choose **QR Pairing**.

![Data source row actions menu with QR Pairing and Code Pairing options.](/docs/data-sources/ai-vision-data-source/04-pairing-menu.png)

3. A QR code opens on the screen.
4. On the Zira AI-Vision device home screen, tap **Pair**.
5. Use the scanner on the device to capture the QR code shown in Zira.

![QR code screen for pairing the AI-Vision device.](/docs/data-sources/ai-vision-data-source/05-qr-code-screen.png)

If you prefer typing a code, use the **OTC** option on the device and choose **Code Pairing** in Zira instead.

Once the QR code is scanned successfully, the device is paired to the data source and ready to continue with the rest of your AI-Vision workflow.

## Continue with guided setup

After the data source exists, you can use **Home AI Assistant** to guide the rest of the workflow.

Ask:

- set up vision device
- what is next for this Vision data source?
- guide me through this device from scratch

The assistant checks the current status and suggests the next step. For a brand-new device, it guides:

1. pairing with QR or OTC
2. mounting confirmation
3. opening **Live View** to verify the camera view
4. opening **Remote Control**
5. recording or scheduling the first video

After videos exist, the assistant can continue the improvement loop:

1. sample frames into a dataset
2. annotate images
3. prepare train and validation splits
4. create a dataset version
5. run training
6. push the model to the device
7. collect validation sessions
8. review the results and move useful examples back into the dataset

## Read next

- [Create a Data Source](/docs/data-sources/create-a-data-source) - Return to the full create flow overview.
- [Data Sources Tabs](/docs/data-sources/channel-data-sources) - Review created data sources in the table.
- [Data Sources Tabs](/docs/data-sources/channel-data-sources) - Continue into the data source workspace from the site or channel list.
- [Home AI Assistant](/docs/ai-assistance/home-ai-assistant) - Use guided setup for an existing Vision data source.
- [Remote Control](/docs/ai-vision/remote-control) - Record the first video after setup.
- [Datasets](/docs/ai-vision/datasets) - Prepare images and annotations for training.


Create an AI-Vision data source, review the template fields, and pair it with a Zira AI-Vision device.

Browse Guides

AI-Pallet QA and Routing

Choose AI-Pallet QA and Routing

Start from the pallet template

Continue in the data source workspace

Continue with guided setup

Read next