If your business gets WhatsApp messages, you already know the problem. Half of them are "what are your hours?" The other half might be real leads. Right now, someone on your team sorts them by hand.
DeskcommCRM just hit GitHub trending this week, picking up over 500 new stars in a single day. It is an open-source, self-hosted CRM built specifically for businesses that sell by chat. The AI agent connects to your WhatsApp number, answers incoming messages from a knowledge base you configure, qualifies leads based on your criteria, moves them through a pipeline, and hands off to a real person when the conversation actually needs one. Your data stays on your own server. No monthly fee.
This is a direct open-source alternative to tools like Kommo, Octadesk, and Intercom, which can run $50 to $150 per seat per month. Each AI agent gets its own knowledge base (RAG), so it answers from your actual business information, not generic guesses. You also get simple WHEN/IF/THEN automations: when a lead uses the word "book," tag them as hot and ping you. The project includes an audited AI-to-human handoff, which is the part that matters most in practice: the AI knows when to stop and get you involved, and there is a log of every time it does.
A small clinic, a renovation contractor, or a shop that gets 30 WhatsApp inquiries a day could realistically have the AI handle 20 of them and only surface the 10 that need a real response.
Try This This Week
- Clone the repo at github.com/melgarafael/DeskcommCRM and spin up the local Docker setup from the README
- Write 10 to 15 FAQ answers covering your most common questions. This becomes the AI agent's starting knowledge base
- Set one automation rule to begin: if a lead mentions "book," "price," or "appointment," tag them as hot and send yourself a notification
- Run test messages through your WhatsApp number before going live so you can see where the handoff triggers
- Set a per-agent spending cap from the start. It keeps AI costs predictable while you tune the knowledge base
The Bigger Picture
Not long ago, this kind of customer-facing AI triage was a feature you paid Intercom $200 a month for. Now it is a GitHub repo, a cheap VPS, and an afternoon. That is the shift build9 keeps watching for: tools that do something genuinely useful, cost nothing to run, and are small enough for one person to own. If your inbox is a daily drain, this one is worth a Saturday.