n8n and Skyvern Cloud: WhatsApp browser agent setup

Uploaded on February 11, 2025, this tutorial explains Ben's n8n and Skyvern Cloud browser automation setup. The demonstrated workflow builder, account settings and APIs belong to the hosted service. It uses Gemini 2.0 Flash for the conversational agent, PostgreSQL for chat memory and OpenAI Whisper to transcribe WhatsApp voice messages. The examples cover job applications, contact form outreach and an Amazon shopping task. The walkthrough uses Skyvern V1 for job workflows and contact form tasks, and V2 for prompt-based shopping. These details describe the dated setup shown in the source.

The agent gathers information before calling a tool, including a resume URL and application preferences. n8n generates search queries, retrieves URLs through SerpAPI and loops over the results. For contact forms, it first fetches each page and checks for input fields before sending it to Skyvern. Ben argues that ordinary API calls and workflow logic should handle the predictable steps, leaving browser interaction for tasks those methods cannot complete.

The WhatsApp connection requires the official Business API and credentials. This is an API-connected AI agent demonstration; the tutorial does not establish an offline or self-hosted deployment. Ben describes browser agents as slow and imperfectly reliable. He reports several minutes per job application, while the shopping example takes roughly eight minutes and adds a cookbook to the cart. It does not complete checkout because he withholds login and payment details.