Documentation: Claude Action (AI Analysis with Anthropic)¶

Overview¶

The Claude Action is an automation node that allows you to use Anthropic's Claude artificial intelligence models to process and analyze content: text, images, audio, and video. It belongs to the AI Models node family and shares the same structure as actions from other providers (Gemini, GPT, Grok).

In IoT and security environments, it is ideal for analyzing camera images with AI when an event occurs (identifying people, vehicles, objects in restricted areas), as well as for processing text with high-quality reasoning.

When to use this action?¶

Use this action when you need to:

Analyze images from cameras (detect people/vehicles, read text, describe scenes).
Analyze video to extract information or detect events.
Process text (summarize, classify, extract data) with Claude models.
Integrate Anthropic capabilities into your automations.

Node Configuration¶

The configuration is divided into two sections, switchable with the top selector: Basic Configuration and Prompt Configuration. It also includes the JSON Editor tab.

Empty basic configuration of the Claude node

Section: Basic Configuration¶

1. API Key *Required¶

Select the Anthropic credential that authenticates access (managed in a centralized and secure way).

2. Resource Type *Required¶

The type of content to process: Text, Image, Audio, or Video.

3. Model *Required¶

The Claude model to use:

Model	Value
Claude 3.5 Sonnet	`claude-3-5-sonnet-20241022`
Claude 3.5 Haiku	`claude-3-5-haiku-20241022`
Claude 3 Opus	`claude-3-opus-20240229`
Claude 3 Sonnet	`claude-3-sonnet-20240229`
Claude 3 Haiku	`claude-3-haiku-20240307`

Basic configuration of the Claude node with Image resource

Section: Prompt Configuration¶

4. Operation *For Image and Video¶

For Image: Analyze Image.
For Video: Analyze Video.

5. Image / Video URLs¶

For analysis, enter the URLs (one per line). Supports template expressions (e.g., {{get_snapshot_node.url}}).

6. Prompt *Required¶

The instruction/question for the model. Supports template expressions.

Prompt configuration of the Claude node

JSON Editor View¶

JSON Editor view of the Claude node

JSON Structure (Input Parameters)¶

{
  "api_key": "",
  "resource": "image",
  "operation": "analyze",
  "model_id": "claude-3-5-sonnet-20241022",
  "image_urls": [
    "{{get_snapshot_node.url}}"
  ],
  "video_urls": [],
  "prompt": "Analyze this camera image and describe whether there are vehicles or unauthorized people in the restricted area."
}

JSON Fields¶

Field	Type	Description
`api_key`	string	Reference to the Anthropic credential (managed securely).
`resource`	string	Resource type: `text`, `image`, `audio`, `video`.
`operation`	string	Operation (image: `analyze`; video: `analyze_video`).
`model_id`	string	Claude model ID (e.g. `claude-3-5-sonnet-20241022`).
`image_urls`	array (string)	URLs of images to analyze.
`video_urls`	array (string)	URLs of videos to analyze.
`prompt`	string	The instruction/question for the model.

Output: Where the node's data comes from¶

The analysis result (text generated by Claude) is available in the node's output and can be used in downstream nodes with {{node_key}}.

Usage Examples¶

Example 1: Detection of unauthorized vehicles/people¶

Use case: When motion is detected in a restricted area, the image is captured and Claude analyzes it to determine whether there are unauthorized vehicles or people.

Resource Type: Image | Model: claude-3-5-sonnet-20241022
Operation: Analyze Image
Image URLs: {{get_snapshot_node.url}}
Prompt: Analyze this camera image and describe whether there are vehicles or unauthorized people in the restricted area.

(see JSON structure above)

Example 2: Classification of a text report¶

Use case: Classify the severity of a received report.

Resource Type: Text
Prompt: Classify the severity (low/medium/high) of the following report and justify it: {{trigger.body.text}}

Validation and Errors¶

Condition	Common cause / fix
Authentication error	The Anthropic credential is invalid or lacks permissions/balance.
URLs not working	Make sure the URLs are publicly accessible (convert internal Docker URLs to their public path).
Usage limit exceeded	Check the limits on your Anthropic account; consider a lighter model (Haiku).

Best Practices¶

Use centralized credentials: Do not write the API key in the node.
Choose the right model: Use Sonnet for a quality/cost balance; Haiku for simple/fast tasks; Opus for maximum capability.
Specific prompts: Ask concrete questions to confirm or dismiss security events.
Chain with snapshot capture: Typical pattern: Get snapshot → Claude (Analyze Image) → condition/notification.