# DeskcommCRM

> Open-source, self-hosted CRM with native AI agents that handle WhatsApp sales conversations, qualify leads and move the funnel.

DeskcommCRM is an open-source, self-hosted sales CRM by Rafael Melgaço (melgarafael) in which AI agents answer customers on WhatsApp, qualify leads and move them through the sales funnel, with actions logged and auditable. It runs on the user's own VPS and is released under the MIT license. The latest GitHub release listed is v1.80.0.

## What It Is

DeskcommCRM ("Desk" + "comm") is a sales operating system for businesses that sell by chat. It combines an inbox, a Kanban funnel, contacts and an AI-agent layer in a single multi-tenant application built on Next.js, TypeScript and Supabase (Postgres, Auth, Storage). The project positions itself as an open alternative to Kommo, Octadesk and Intercom.

## How the AI agents work

Agents use per-tenant RAG over the company's own knowledge base, run skills during conversations, keep organizational memory and hand off to a human with a summary when appropriate. According to the project, every outgoing message passes seven checks in a fixed order (unsubscribe, LGPD, anti-ban, text variation, deterministic promise, semantic promise and automation notice), and each evaluation, including blocked messages, is recorded. Leads can move between funnel stages automatically with the reason logged, and a Radar view classifies stalled demands by risk. Follow-ups can be scheduled to revive cooled conversations. Further modules include AI evolution screens and proposals that require human approval.

## Channels, automation and configuration

WhatsApp connects via QR code (WAHA, multi-number, with throttle, jitter and time-window anti-ban controls) or the official Meta Cloud API with synced templates. Inbound webhook sources accept leads from landing pages, Zapier or n8n, and WHEN/IF/THEN automations can add tags, move leads, assign agents, send WhatsApp messages or call outbound webhooks. The funnel vocabulary is configurable per niche (e-commerce, clinics, real estate, services), a Nuvemshop integration exists, and the AI provider (OpenRouter, Anthropic or OpenAI) can be chosen at install and changed later. Governance features include RBAC, audited assignment and transfer, queue rotation and routing by intent, with LGPD-oriented multi-tenancy using row-level security.

## Setup path

Installation is done on a VPS with a single script (hostgator-setup-kit/install.sh) that checks dependencies, generates secrets, applies the database schema, creates the first admin, starts the containers with automatic HTTPS and installs the cron that drives automations. 4 GB of RAM is recommended; the stack starts with 2 GB, with seven containers and roughly 150 MB per WhatsApp session. A Supabase project, a domain and an AI provider key are needed. Updates run via update.sh or an in-app screen, with a database backup beforehand. Assistant guides for tools such as Claude Code, Codex and Cursor can walk through installation.

## Agencies and white-label

Two environment variables change the name and logo across the interface. The project states that colors and fonts require changing the design system and that branding is per installation, not per organization.

## Features
- AI agents that answer on WhatsApp, qualify leads and move the funnel
- Per-tenant RAG knowledge base
- Seven pre-send checks with durable audit records
- AI-to-human handoff with context summary
- Radar for leads at risk of going cold
- Adaptive follow-ups
- Configurable funnel vocabulary per niche
- WhatsApp via QR code (WAHA) or Meta Cloud API
- Inbound webhooks and WHEN/IF/THEN automations
- RBAC, audited assignment and routing
- Multi-tenant isolation with RLS and LGPD features
- One-command install and in-app updates with backup
- White-label name and logo via environment variables
- Choice of AI provider (OpenRouter, Anthropic, OpenAI)

## Integrations
WhatsApp (WAHA), WhatsApp Meta Cloud API, Supabase, OpenRouter, Anthropic, OpenAI, Nuvemshop, Zapier, n8n, Upstash Redis, Sentry, Docker

## Platforms
WINDOWS, MACOS, LINUX, WEB, API

## Pricing
Open Source

## Version
v1.80.0

## Links
- Website: https://deskcomm.com.br
- Documentation: https://github.com/melgarafael/DeskcommCRM/blob/main/docs/SETUP.md
- Repository: https://github.com/melgarafael/DeskcommCRM
- EveryDev.ai: https://www.everydev.ai/tools/deskcommcrm