Anthropic Claude

You can call the Anthropic's Claude API from a Workflow.

About this block

The Anthropic Claude block lets you call the Anthropic Claude API from a Workflow.

You can use this block for any multimodal task supported by Workflows.

You can then return the result from the API directly in your Workflow output.

You can also use the result as an input to other blocks. For example, you can ask Claude a yes/no question, process the result with the VLM as Classifier block, then use the information to control the flow of your Workflow with a Continue-If block.

To use this block, you will need an Anthropic API key. Learn how to create an Anthropic API key.

The Single-Label Classification Model block.

What you can send into this block

You can run the Object Detection Model block on:

1. An image that you send to your Workflow

2. A video frame that you send to your Workflow

3. A cropped region of an image that is calculated in your Workflow (i.e. with a Crop block)

How to configure the block

When you set up this block, you will need to provide:

• A task type.

• A prompt to send to Claude.

• Your Anthropic API key

Open Prompt is the default task type. This lets you send your own prompt to Claude.

You can choose between other task types such as single-label or multi-label classification. These tasks are configured to return data in a format that other Workflow blocks will understand. This is ideal if you plan to use Claude with other blocks in your Workflow.

You can also choose what model version to use from the Additional Properties section:

Select a model version.

What this block returns

Use cases

This block is useful for any Workflow from which you want to call a cloud vision language model (VLM).

If you need a VLM that runs on your own hardware, we recommend trying the Qwen-VL 2.5 block.

Predictions format

Expand the card below to see the JSON data returned by this block.

Predictions format

For an Open Prompt request, this block returns data in the following format:

[
  {
    "output":  , 
    "anthropic_claude": {
      "output": "The image shows a user interface for an AI model task setup. In the interface, there's a \"Task Type\" section where \"Open Prompt\" is selected. Below that is a \"Prompt\" field containing the text: \"What is the dock number? Return like A1.\"

There is also an \"API Key\" field which appears to be filled with a masked value (shown as dots for security).

I don't see any dock number in the image itself. The text in the prompt field appears to be instructions for an AI model to identify a dock number and format the response in a specific way (like \"A1\"), but the actual dock number is not present in the image you've shared.",
      "classes": null
    }
  }
]

Deployment considerations

To use this block, you will need an internet connection. This means you cannot deploy this block on the edge unless you have an internet connection.