Concierge
Production-grade WhatsApp orchestration daemon for OpenClaw agent workflows.
v2.0.0 · Elixir/OTP
Production-grade WhatsApp orchestration daemon for OpenClaw agent workflows.
v2.0.0 · Elixir/OTPThe missing orchestration layer between WhatsApp and your AI agents.
Every WhatsApp contact gets its own Elixir process with dedicated state, context, and identity. Crashes are contained.
Sessions run inside OpenClaw Docker containers with restricted workspace access and no tools by default.
Detects manual replies and auto-pauses the bot, preventing conversations from being hijacked by AI.
Three quick texts become one AI response. Smart wait windows reduce noise and improve response quality.
Responds to missed messages on startup or on schedule. Never leaves a customer hanging after downtime.
All DB writes are serialized through dedicated GenServers. No lock contention, no corruption, no surprises.
A clean data flow from WhatsApp to AI and back, powered by OTP supervision.
Concierge.Supervisor
├── ChatRegistry
├── ConfigServer
├── WacliDB
├── Store
├── AgentManager
├── SyncController
├── SandboxCleaner
├── ChatSupervisor
└── Monitor
Everything you need for production WhatsApp AI workflows.
Up and running in two steps.
brew tap arvorco/tap
brew install concierge --HEAD
cp config.example.json config.json
bin/concierged
git clone https://github.com/arvorco/concierge
cd concierge
mix deps.get
cp config.example.json config.json
mix run --no-halt