DeskcommCRM: configure WhatsApp agents with Claude Code

This Portuguese-language tutorial covers configuring an existing DeskcommCRM installation with Claude Code. The presenter describes DeskcommCRM as an open source sales system with WhatsApp messaging, CRM funnels and automated follow-ups. The focus is client configuration rather than installing the CRM.

The setup requires a VPS with the CRM already installed and an AI CLI on the user's computer. The presenter recommends 4 GB of RAM for starting out, says 2 GB is insufficient for the system shown, and suggests 8 GB for a larger setup with a separate database. Installation guidance distinguishes Windows PowerShell from macOS and Linux, and the demonstrated Claude login uses a subscription account.

Claude Code connects to the server over SSH and studies the running system and business context. For a spoken browser walkthrough, the presenter keeps Claude on his computer and connects the Claude extension in Chrome. A disconnected extension interrupts the demonstration until he authorizes access.

The configuration example asks for a sales AI agent that qualifies incoming leads, plus meeting reminders and follow-ups for missed meetings. The presenter directs Claude to use SQL rather than browser clicks, which he says is faster. He then reviews the generated agent instructions, RAG knowledge documents and workflow blocks. He also describes keeping task questions in documents while Claude works. This is a self-hosted CRM workflow; the tutorial does not demonstrate local model inference.