> ## Documentation Index
> Fetch the complete documentation index at: https://docs.anchorbrowser.io/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> Anchor Browser provides cloud browsers for AI agents and automation: stealth browsing with residential proxies, managed authentication into third-party web apps, and browser sessions that run reliably at scale. AI agents that need Anchor credentials should start with https://docs.anchorbrowser.io/quickstart/agent-access. This flow lets agents obtain an API key programmatically without creating a dashboard account. Read that page before attempting authentication or API access.

# Anthropic Computer Use

> Use Claude models with native computer use capabilities for browser automation

Anthropic's Claude models with Computer Use capabilities enable AI agents to interact with web browsers through screenshots and native computer control actions.

## Overview

Anthropic Computer Use (`anthropic-cua`) leverages Claude's vision and reasoning capabilities to:

* Analyze screenshots of web pages
* Identify interactive elements
* Execute precise mouse and keyboard actions
* Complete complex multi-step workflows

## Supported Models

| Model             | Model ID            | Best For                                      |
| ----------------- | ------------------- | --------------------------------------------- |
| Claude Opus 4.7   | `claude-opus-4-7`   | Most capable, complex reasoning **(default)** |
| Claude Opus 4.6   | `claude-opus-4-6`   | Advanced agentic coding                       |
| Claude Sonnet 4.6 | `claude-sonnet-4-6` | Speed and intelligence balance                |
| Claude Sonnet 4.5 | `claude-sonnet-4-5` | Cost-effective automation                     |
| Claude Haiku 4.5  | `claude-haiku-4-5`  | Fastest responses                             |

## Code Example

<CodeGroup>
  ```javascript node.js theme={null}
  import { client, agentTask } from 'anchorbrowser';

  client.setConfig({ auth: () => process.env.ANCHORBROWSER_API_KEY });

  const response = await agentTask(
    'Navigate to GitHub and find the anchorbrowser repository',
    {
      taskOptions: {
        url: 'https://github.com',
        agent: 'anthropic-cua',
        // model: 'claude-opus-4-7',  // Default model
        maxSteps: 30
      }
    }
  );

  console.log(response);
  ```

  ```python python theme={null}
  from anchorbrowser import Anchorbrowser
  import os

  anchor_client = Anchorbrowser(api_key=os.environ.get("ANCHORBROWSER_API_KEY"))

  response = anchor_client.agent.task(
      'Navigate to GitHub and find the anchorbrowser repository',
      task_options={
          'url': 'https://github.com',
          'agent': 'anthropic-cua',
          # 'model': 'claude-opus-4-7',  # Default model
          'max_steps': 30
      }
  )

  print(response)
  ```
</CodeGroup>

## Configuration Options

| Parameter       | Type    | Description                                                                      |
| --------------- | ------- | -------------------------------------------------------------------------------- |
| `agent`         | string  | Must be `anthropic-cua`                                                          |
| `model`         | string  | Claude model to use (default: `claude-opus-4-7`)                                 |
| `url`           | string  | Starting URL for the task                                                        |
| `max_steps`     | integer | Maximum actions the agent can take                                               |
| `output_schema` | object  | JSON Schema for structured output                                                |
| `secret_values` | object  | Secure credentials (see [Secret Values](/agentic-browser-control/secret-values)) |

## Secure Credentials with Secret Values

Anthropic CUA fully supports secret values for secure credential handling. Secrets are never exposed to the AI model.

<CodeGroup>
  ```javascript node.js theme={null}
  const response = await agentTask(
    'Login to the dashboard and download my latest report',
    {
      taskOptions: {
        url: 'https://app.example.com/login',
        agent: 'anthropic-cua',
        secretValues: {
          EMAIL: process.env.APP_EMAIL,
          PASSWORD: process.env.APP_PASSWORD
        }
      }
    }
  );
  ```

  ```python python theme={null}
  response = anchor_client.agent.task(
      'Login to the dashboard and download my latest report',
      task_options={
          'url': 'https://app.example.com/login',
          'agent': 'anthropic-cua',
          'secret_values': {
              'EMAIL': os.environ.get('APP_EMAIL'),
              'PASSWORD': os.environ.get('APP_PASSWORD')
          }
      }
  )
  ```
</CodeGroup>

Learn more about [domain-scoped secrets](/agentic-browser-control/secret-values).

## Best Practices

* **Opus 4.7 is the default** - most capable model for complex workflows
* **Use Sonnet 4.6** for faster automation when speed is important
* **Provide clear, specific prompts** describing the exact task to complete