Anchor Browser icon

Anchor Browser

Interact with Anchor Browser API for browser automation and control

Actions35

Tool → Perform Web Task

AI-generated

Summary

Execute an AI-powered web task via Anchor Browser API, optionally within an existing session or at a specified URL.

Inputs

  • prompt (required) — The task to be autonomously completed. When using Output Schema, be specific about what data to extract.
  • url (required) — The URL to perform the task on.
  • sessionId — The ID of the browser session. Optional - if not provided, a new session will be created.
  • agent — The AI agent to use for task completion. Options: Anthropic CUA, Browser Use, Gemini Computer Use, OpenAI CUA.
  • async — Whether to run the task asynchronously. If true, the task will be run asynchronously and the response will include a workflow ID.
  • detectElements — Whether to detect elements during task execution.
  • extendedSystemMessage — Extended system message to provide additional context for the AI agent.
  • highlightElements — Whether to highlight elements during task execution.
  • humanIntervention — Whether to allow human intervention during task execution.
  • maxSteps — Maximum number of steps for task completion.
  • model — The specific model to use for task completion.
  • outputSchema — JSON Schema defining the expected structure of the output data.
  • provider — The AI provider to use for task completion. Options: Azure, Gemini, Groq, OpenAI, xAI.
  • secretValues — Secret values (key-value pairs) to be used during task execution. Keys must start with ANCHOR_.
  • useVision — Whether to use vision capabilities for the AI agent.

Output shape

If the task completes synchronously, returns the processed task result data (object or parsed from API response). If run asynchronously (async=true), returns an object containing status and workflowId, which can be queried later using the 'Get Perform Web Task Status' operation. The returned data shape depends on the API response: if a workflowId is present, the output includes status and workflowId; if a result is present, it contains the extracted data per the outputSchema; otherwise the raw API data is returned.

The API endpoint is POST /v1/tools/perform-web-task. Either the sessionId or URL must be provided; prompt is required. All other parameters are optional and mapped to API body fields as defined in the node source.

Links

Discussion