DeskcommCRM: Self-hosted WhatsApp AI sales CRM
A self-hosted AI CRM for WhatsApp sales teams, combining conversations, funnels, and RAG agents.
GitHub melgarafael/DeskcommCRM Updated 2026-09-12 Branch main Stars 1.3K Forks 496
Next.js 16 WhatsApp sales WAHA Plus Supabase Self-hosted VPS

🧭 Decision Guide

Why trending now: The README simultaneously emphasizes WhatsApp AI agents, self-hosted VPS deployment, MCP, LGPD, and alternatives to Kommo/Octadesk/Intercom. Release data shows 152 new stars and a daily Trending appearance on 2026-09-12, but the material is insufficient to determine the specific cause.

Try it if you

  • You need a self-hosted WhatsApp CRM on a VPS and want AI agents involved in support and sales.
    README sections “O que é” and “Stack”: the CRM runs on a VPS, WAHA Plus and Meta Cloud API provide WhatsApp, and AI agents can attend, qualify, and sell.
  • Your team uses Supabase and needs RLS, vector, and tenant-isolated RAG.
    README “Stack”: the database is Supabase (Postgres + RLS + vector); “O que é” mentions agents de IA com RAG por tenant.
  • You want to use OpenRouter, Anthropic, OpenAI, or Google and change AI providers through the UI.
    README “Stack”: Vercel AI SDK v7 supports these providers; the installer asks which one to use, and it can later be changed in the UI.
  • You need Inbox, Kanban, Follow-ups, Audit Log, and 196 API route handlers.
    README “O que você opera” lists the relevant screens; “Estrutura” states that the REST API has 196 route handlers.

Skip it if you

  • You only want a hosted SaaS and do not want to maintain a VPS, Docker, WAHA Plus, or Supabase.
    README “Stack” specifies VPS com Docker as the hosting model; the installation material centers on self-hosting and VPS deployment.
  • You need standard Supabase migrations to initialize the database from scratch.
    README “Desenvolvimento” explicitly says migrations 0001-0009 and 0013 are stubs; `supabase db push` can pass while leaving an empty database, so baseline.sql must be used.
  • Your business does not use WhatsApp and does not need WAHA or Meta Cloud API in the CRM.
    The README title and core description focus on WhatsApp; “Stack” lists WAHA Plus and Meta Cloud API as the central channel components.

Requirements

  • Production: VPS with Docker; the README also identifies HostGator/SP as the partner deployment environment.
  • Development: `nvm use` (Node 22), `pnpm`, Docker, and a Supabase project.
  • A new Supabase project must enable the `vector`, `citext`, and `pg_trgm` extensions.
  • `docs/SETUP.md` covers Supabase, WAHA, AI providers, Upstash, Sentry, Resend, and Nuvemshop integrations; the README estimates approximately 60–90 minutes.

First step (verbatim from README)

curl -fsSL https://raw.githubusercontent.com/melgarafael/DeskcommCRM/main/hostgator-setup-kit/comecar.sh | bash

Watch out

  • Do not run only `supabase db push`; the README explicitly says it can leave an empty database.
    README “Desenvolvimento”: the migrations are stubs and the schema is in `supabase/baseline.sql`.
  • Without `vector`, applying the baseline can fail with `type public.vector does not exist`.
    README “Desenvolvimento” gives this exact error and the extension-creation command.
  • Local WAHA is optional for development, but WhatsApp integration still involves WAHA Plus or Meta Cloud API.
    README “Desenvolvimento” marks `docker compose up -d` as WAHA local (optional); “Stack” lists both WhatsApp channels.
  • Do not submit security vulnerabilities as public Issues; use the private vulnerability report described in SECURITY.md.
    README “Reportando bugs” explicitly says not to open a public issue for security vulnerabilities.

Alternatives

  • Kommo：When you want to evaluate the commercial CRM alternative explicitly named by the README instead of deploying DeskcommCRM’s VPS, Docker, and WhatsApp components.
    README “O que é”
  • Octadesk：When you prefer the ready-made commercial product named by the README and do not want to maintain Supabase, WAHA Plus, and workers yourself.
    README “O que é”
  • Intercom：When you want to compare a mature commercial customer-communication platform rather than an MIT-licensed self-hosted WhatsApp CRM.
    README “O que é”

Not stated in the README

  • The omitted sections do not provide minimum VPS CPU, memory, disk, or bandwidth requirements.
  • The material does not state actual costs or quotas for WAHA Plus, Meta Cloud API, Supabase, Upstash, Sentry, or AI providers.
  • The material does not state supported WhatsApp account scale, concurrent-session limits, or per-tenant capacity.
  • The material does not describe the specific changes in v1.19.0 compared with the previous release.
  • The material does not provide the submission cadence or long-term maintenance plan associated with 10 contributors and 10 recent commits.
  • The material does not identify the direct source of the 152 stars added on Trending day; it cannot confirm whether AI, WhatsApp, self-hosting, the HostGator partnership, or another factor caused the attention.

💡 Deep Analysis

6
No I contribute with Node 22 in a Next.js 16, React 19, and TypeScript 6 environment. Can I initialize DeskcommCRM through the normal Supabase migration workflow, or must I use the project's baseline.sql?
For: An open-source contributor maintaining Next.js 16, Node 22, and Supabase, who needs to run WAHA locally, initialize the database, and modify TypeScript code

No, not if you follow the ordinary supabase db push workflow; the README explicitly requires enabling extensions first and then applying supabase/baseline.sql.

  • The development environment requires Node 22 and pnpm; the frontend uses Next.js 16 App Router, React 19, and strict TypeScript.
  • The README states that migrations 0001–0009 and 0013 are SELECT 1; stubs. The migration chain does not build the real schema from scratch, and supabase db push may “pass” while leaving an empty database.
  • A new Supabase project must first enable vector, citext, and pg_trgm; otherwise the baseline can fail with type public.vector does not exist.
  • Local WAHA is optional. The documented development path includes Docker Compose, applying the baseline, and pnpm dev, so contributors must follow the project-specific initialization path.
  • README, “Desenvolvimento”: `nvm use # Node 22` and `npm install -g pnpm && pnpm install`
  • README, “Desenvolvimento”: migrations 0001–0009 and 0013 are stubs; `supabase db push` leaves an empty database
  • README, “Desenvolvimento”: enable vector, citext, and pg_trgm before applying `supabase/baseline.sql`
  • README, “Desenvolvimento”: `docker compose up -d` and `pnpm dev`
git clone https://github.com/melgarafael/DeskcommCRM.git
Not stated in the README：The README does not describe the formal migration process between baseline.sql and later application versions.；The README does not specify minimum Docker or operating-system versions for local WAHA, Supabase, and Node 22 development.
Yes I do not want vendor lock-in and need to switch among OpenRouter, Anthropic, OpenAI, and Google while tracking agent executions, usage, and budgets. Does DeskcommCRM meet this constraint?
For: A technical lead at a self-hosting company who wants control over AI vendors, model costs, and business data, and needs to switch among OpenRouter, Anthropic, OpenAI, and Google

Yes, because AI providers and credentials are configurable in the management interface, and the project treats executions, usage, and budgets as first-class parts of AI-agent management.

  • The README’s AI stack uses Vercel AI SDK v7 and lists OpenRouter, Anthropic, OpenAI, and Google. The installer asks for a provider, and the provider can later be changed through the UI.
  • The AI Agent area includes Providers, Credentials, Executions, and Usage and budget, matching the needs for provider configuration, execution records, and cost control.
  • Agents also support RAG, Memory, Skills, Cases, Alerts, Proposals, and human handoff, so this is more than a fixed-model Q&A endpoint.
  • External model latency, context limits, and inference charges remain your responsibility; the project insight also calls out RAG noise and ongoing reasoning cost as separate evaluation areas.
  • README, “Stack”: Vercel AI SDK v7 with OpenRouter, Anthropic, OpenAI, and Google; the installer asks for a provider and it can later be changed in the UI
  • README, “O que você opera”: Providers, Credentials, Executions, and Usage and budget
  • README, “O que é”: AI agents use RAG, advance sales stages, trigger automations, and hand off to humans
  • Project insight: the AI layer is provider-replaceable and includes usage, budget, and execution records
cp .env.example .env.local  # guia completo em docs/SETUP.md
Not stated in the README：The README does not list specific model names, default models, or a compatibility matrix for switching providers.；The README does not state whether exceeding a budget blocks execution, downgrades the model, or only records an alert.
Yes We sell mainly through WhatsApp and want human agents and AI to collaborate in one Inbox while keeping the CRM, follow-ups, and customer data on our own Docker VPS. Is DeskcommCRM suitable for us?
For: An SMB owner whose sales happen mainly on WhatsApp, who wants human agents and AI in one inbox, and plans to self-host on a Docker VPS

Yes, because it is designed specifically for WhatsApp sales, human–AI collaboration, and self-hosted deployment.

  • The README describes Inbox as “conversas de WhatsApp, você e a IA lado a lado,” and also includes Kanban, contacts, follow-ups, Radar, and sales-performance screens.
  • WhatsApp supports both WAHA Plus QR connections and the Meta Cloud API; the README positions the former for getting started quickly and the latter for scale.
  • The installer deploys the app, WhatsApp component, and database on a VPS, while AI providers can be configured through OpenRouter, Anthropic, OpenAI, or Google.
  • The trade-off is operational ownership: the company must manage the VPS, database, channel, AI costs, and security. The README estimates roughly 60–90 minutes for full integration setup.
  • README, “O que você opera”: Inbox with “conversas de WhatsApp, você e a IA lado a lado”
  • README, “Stack”: WAHA Plus (NOWEB engine) + Meta Cloud API; VPS with Docker
  • README: `docs/SETUP.md` covers all integrations and estimates about 60–90 minutes
  • Project insight: combines WhatsApp Inbox, CRM, AI agents, and follow-ups into a self-hosted sales OS
curl -fsSL https://raw.githubusercontent.com/melgarafael/DeskcommCRM/main/hostgator-setup-kit/comecar.sh | bash
Not stated in the README：The README does not specify VPS sizing or performance limits for a particular message volume, agent concurrency, or contact count.；The README does not describe the exact process for migrating an existing WhatsApp number to the Meta Cloud API.
It depends We currently use WAHA Plus QR mode to validate WhatsApp sales and now want to expand operations while reducing unofficial-channel risk. Does DeskcommCRM support moving to the Meta Cloud API?
For: A sales operations lead who validated WhatsApp workflows with WAHA QR mode and now plans to move to the official Meta Cloud API channel for broader operations

It depends: the project supports both WAHA Plus and the Meta Cloud API, but switching from QR mode still requires checking the number, templates, webhooks, and messaging rules.

  • The README’s Stack lists WAHA Plus (engine NOWEB) + Meta Cloud API, describing QR as the quick-start option and the official Meta channel as the scaling path.
  • The Connections screen includes connection health, reconnection, and template management, showing that the product addresses channel operations.
  • The project insight warns that WAHA can face session disconnects, re-authentication, WhatsApp policy changes, and unofficial-channel risk; production and compliance scenarios should prioritize the Meta Cloud API.
  • However, the README provides no migration script, number-migration procedure, or Meta approval checklist. Supporting both channels does not mean migration is automatic.
  • README, “Stack”: WAHA Plus (NOWEB engine) + Meta Cloud API; QR for getting started quickly and the official channel for scale
  • README, “O que você opera”: Connections with QR or official Meta channel, health, reconnection, and templates
  • Project insight: WAHA may face disconnections, re-authentication, policy changes, and unofficial-channel risk
  • Project insight: production-scale and compliance scenarios should prioritize the Meta Cloud API
curl -fsSL https://raw.githubusercontent.com/melgarafael/DeskcommCRM/main/hostgator-setup-kit/comecar.sh | bash
Not stated in the README：The README does not explain whether a WAHA number can be converted to a Meta Cloud API number or how message history would be retained.；The README does not explain how Meta template approval, messaging windows, and official API fees map to the product’s templates or budget screens.
It depends I need to maintain a multi-tenant CRM on Next.js 16, React 19, and strict TypeScript, while allowing AI to move leads, run follow-ups, and read tenant knowledge through MCP. Can DeskcommCRM serve as the foundation?
For: A TypeScript developer serving multiple organizations or clients from one CRM, relying on Supabase Postgres/RLS for tenant isolation and wanting AI to operate the CRM through MCP

It depends: the platform capabilities fit, but using it as a production foundation depends on your ability to own RLS, MCP authorization, and business customization risks.

  • The stack includes Next.js 16 App Router, React 19, strict TypeScript 6, and Supabase Postgres with RLS and vector support, which fits extending CRM and RAG in one codebase.
  • The README states that AI agents use tenant-scoped RAG and can operate the CRM through MCP. It also exposes organizations, teams, API tokens, Audit Log, and LGPD screens.
  • However, the project insight states that tenant security depends on correct RLS policies for every table, server authorization, MCP tool permissions, and token management; mistakes can expose data across organizations.
  • The system also relies on Supabase Auth, Realtime, Storage, and Postgres extensions, so moving to another database would require additional work.
  • README, “Stack”: Next.js 16 App Router, React 19, TypeScript 6 strict, and Supabase with Postgres, RLS, and vector
  • README, “O que é”: tenant-scoped RAG agents and the full CRM exposed through MCP
  • README, “O que você opera”: Organization, Team, Audit Log, LGPD, API Tokens, and Security
  • Project insight: multi-tenant security depends on RLS, server authorization, MCP tool permissions, and API token management
git clone https://github.com/melgarafael/DeskcommCRM.git
Not stated in the README：The README does not list the complete RLS policy for every business table or a default permission matrix for each MCP tool.；The README does not state maximum organization counts, concurrent AI executions, or Supabase resource baselines for multi-tenant deployments.
It depends We use Nuvemshop, complete sales mainly through WhatsApp, and require LGPD screens, Audit Log, MFA, and private WhatsApp media storage. Is DeskcommCRM sufficient as the CRM foundation?
For: A Brazilian e-commerce technical lead who needs to connect Nuvemshop, sales pipelines, and WhatsApp automation while requiring LGPD, audit logs, and private media storage

It depends: the product covers the system capabilities you listed, but those screens alone do not prove that all LGPD obligations are fulfilled.

  • The README explicitly lists Nuvemshop, Webhooks, WhatsApp Inbox, Kanban, and automations, making it a plausible business entry point for e-commerce chat sales.
  • The Organization and Security areas include LGPD, Audit Log, MFA, recovery codes, sessions, and API Tokens. The Stack also specifies a private whatsapp-media Supabase Storage bucket with signed URLs.
  • The project insight says the system provides audit and privacy-related capabilities, but the company remains responsible for retention, deletion requests, consent, supplier agreements, and cross-border processing.
  • It can therefore provide a compliance implementation base, but it cannot replace the company’s legal procedures, data classification, or supplier review.
  • README, “O que você opera”: Nuvemshop, Webhooks, LGPD, Audit Log, MFA, and API Tokens
  • README, “Stack”: Supabase Storage with signed URLs and private bucket `whatsapp-media`
  • README, “O que é”: the project originated as an e-commerce CRM and now supports stores and WhatsApp sales
  • Project insight: LGPD features exist, but retention, deletion, consent, supplier agreements, and cross-border processing remain the deployer’s responsibility
cp .env.example .env.local  # guia completo em docs/SETUP.md
Not stated in the README：The README does not specify which Nuvemshop events, field mappings, or synchronization directions are supported.；The README does not state whether the LGPD screens automatically execute deletion, export, and consent recording, or merely provide management entry points.；The README does not specify the default retention period or backup policy for private media.

✨ Highlights

  • WAHA Plus connects WhatsApp with AI agents and human collaboration
  • Supabase Postgres RLS and vector support multi-tenant RAG
  • The VPS setup kit can deploy the full CRM with one curl command
  • The development environment requires Node 22, Docker, and Supabase extensions

🔧 Engineering

  • Inbox, Kanban, and Radar manage WhatsApp conversations and sales funnels
  • Vercel AI SDK v7 integrates OpenRouter, Anthropic, OpenAI, and Google
  • MCP exposes CRM operations for agents to handle leads, automations, and handoffs

⚠️ Risks

  • Migrations 0001-0009 and 0013 are stubs; baseline.sql is required to create the schema
  • New Supabase projects must enable vector, citext, and pg_trgm first
  • SETUP.md estimates 60–90 minutes to configure Supabase, WAHA, and other integrations
  • Production deployment depends on a VPS, Docker, WAHA Plus, and Meta Cloud API

👥 For who?

  • Teams needing WhatsApp sales workflows and multi-tenant RAG
  • Full-stack developers using Next.js, TypeScript, and Supabase
  • Businesses willing to self-host CRM and WhatsApp on a Docker VPS