I'm using GoHighLevel (GHL) with a Conversational AI bot that handles incoming WhatsApp messages, answers customer questions, and qualifies leads before handing them over to my sales team. What I'm trying to build: An internal AI assistant that I can chat with in natural language to analyze conversations and CRM data across my entire GHL sub-account. For example, I want to ask: - How many customers contacted us today? - How many were qualified leads versus irrelevant inquiries? - How many requested actual quotations? - Which leads are waiting for a response or follow-up? - Which conversations should my sales team prioritize? - Can you summarize today's conversations and identify potential lost sales opportunities? What I've already tried: 1. Claude in Chrome: I've been using Claude to read and summarize conversations directly from the GHL interface. It works, but it's inefficient and not scalable. 2. Claude + GoHighLevel MCP: I also connected HighLevel to Claude through MCP. However, the connection keeps disconnecting every other day without any clear explanation. I have to reconnect it manually, and I don't always realize when it has stopped working. This makes it unreliable for daily business operations. What I'm considering: - GoHighLevel Managed AI Agents - GHL MCP Server integration - Claude or ChatGPT connected directly to GHL - Custom API integrations or automation workflows My main questions: 1. Has anyone successfully built an AI agent that can access and analyze conversations across an entire GHL sub-account, rather than just individual contacts? 2. Has anyone experienced frequent MCP disconnections between Claude and GHL? If so, how did you fix them? 3. Can GHL Managed Agents handle this type of internal reporting and analysis? 4. What's the most reliable architecture for keeping an AI assistant connected to GHL without constantly reconnecting or manually exporting data? My ideal setup: I want to open an AI chat, ask questions about my WhatsApp conversations and sales pipeline, and receive accurate, actionable answers based on real CRM data.