WhatsApp support with AI that delivers.
Install in one command
$ curl -fsSL https://app.fazer.ai/api/agents/install.sh | bashAI agents that run on top of your Chatwoot and reply like a person: they listen to voice notes, read images, book appointments, move deals through the pipeline, and hand off to your team when a human is needed. All on your own infrastructure.
Handles chats like a human
It listens to voice notes, reads images, and replies by voice or text. It waits for the customer to finish typing, shows “typing…,” and answers at a conversational pace.
Acts, doesn't just reply
It checks your knowledge base, uses your tools and integrations, books appointments, moves deals through your pipeline, and brings in a human when needed.
Up in one command
One line in the terminal, and an AI assistant installs everything, checks that it works, and hands you an agent that's already answering. On your own infrastructure, with your data.
Watch the agent at work.
Five everyday scenes in faithful miniatures of the console — they play on their own and the buttons really work. This is how your customers will be served.
A human pace
Your customer fires off three messages and a voice note. The agent waits for them to finish, transcribes, shows “typing…” and replies in short bubbles — by text or by voice, at a human pace.
- Batches bursts of messages before replying
- Transcribes voice notes, reads images and documents
- Replies in natural voice, as a WhatsApp voice note
- Reacts with emoji — and knows when not to reply
Acts, doesn't just reply
Charging, scheduling, moving the pipeline: the agent uses real tools. It creates the PIX charge via Asaas and, when the payment lands, the webhook wakes it up — in the right conversation.
- Ready-made toolpacks: Asaas, Google Calendar, Google Drive
- Your own HTTP tools and external MCP servers
- The pipeline kanban follows the conversation
- External events wake the agent on their own
Answers with sources, learns with approval
Answers come from your knowledge base, source included. And when the agent learns something new in a conversation, it suggests it — nothing enters the base without your review.
- Knowledge bases (RAG) in your own database
- PDF, DOCX, TXT, MD and CSV, indexed in one click
- Agent suggestions go through an approval queue
AI and humans on the same team
Sensitive case? The agent wraps up with the customer, leaves a private summary and hands off to the right team. When the human is done, they hand it back — and the AI picks up with full context.
- Chatwoot routing, a fixed team, or let the AI choose
- Private summary before every handoff
- Human takes over? The bot goes quiet instantly
- Hand back to the AI, with re-engagement
Proactive at the right time
Customer went quiet? Follow-ups in stages, only within allowed hours — and the agent decides when insisting is no longer worth it. Appointment reminders go out on their own.
- Staged sequence, with instructions per stage
- Respects your business hours and the 24-hour window
- Reminders can ask to confirm attendance
A complete operations hub.
Automation funnel and cost per conversation, live conversations, the agent editor, knowledge with approvals and a log of every step. Browse around — it is a faithful miniature with sample data.
Click the side menu to explore the screens — Dashboard, Conversations, Agents, Components and Logs.
Built for real operations.
What else comes in the box — all self-hosted, all included in the Free edition.
Real guardrails
A moderation agent reviews input and output: instruction adherence, competitors, tone. You define what happens on a violation.
WhatsApp → website chat
Use the official number just as a front door and take the lead to your website chat — no per-message WhatsApp fees.
Any model, even local
OpenAI, Anthropic, Google, DeepSeek, OpenRouter, or models on your own machine (Ollama, vLLM) via a compatible endpoint.
Omnichannel via Chatwoot
Born on WhatsApp, it also covers Instagram, Telegram, email and website chat — one agent for many inboxes.
Business hours
Weekly windows with time zone. Outside hours the agent stays silent — and follow-ups wait for the next window.
Quotes as PDFs
Items, prices and terms in a PDF behind a secure link, generated right in the conversation.
Playground and test mode
Test edits live without touching real data — and run in production only for whoever types /teste.
Signed webhooks
Conversation created, handed off, card moved: events with HMAC signing, retries and a dead-letter queue.
REST API + MCP server
Operate everything from Claude, Cursor or any MCP client, with OAuth 2.1, scopes and dry-run by default.
Credential vault
Encrypted, write-only keys with connection tests. No secrets in plain text.
Costs under control
Real LLM cost per agent, per model and per inbox, with Langfuse plugged in.
Multi-tenant and white-label
Several isolated clients in one deploy and a console with your brand — in the Pro edition, for agencies.
From the first command to the agent answering.
No 40-step tutorial. You paste the command, and an AI agent in your terminal handles the entire install.
Paste the line in your terminal
Just one command, on macOS, Linux, or Windows.
An AI agent takes over
It installs the stack on your server, connects your tools, and sets everything up together with you.
Validates and delivers
It tests a full conversation and leaves your agent ready to answer on WhatsApp.
$ curl -fsSL https://app.fazer.ai/api/agents/install.sh | bashFree, with everything included
Apache 2.0 license. Every feature, for one business, on your infrastructure and with no lock-in. It's the edition you install with the command above.
View the code on GitHubMany clients, one operation
Serve multiple brands from one deployment, each fully isolated, and hand every client a dashboard with your own branding (white-label). Built for anyone running support for others.
Explore the Pro editionStill have questions?
Is fazer.ai agents free?
Yes. The Free edition is open source (Apache 2.0) and includes every feature, for one business. The Pro edition, aimed at agencies, adds multiple clients in a single deploy and your own branding (white-label).
Do I need to know how to code to install it?
No. You paste a single line in your terminal and an AI assistant runs the installation: it brings up the stack on your server and validates that everything works.
Does it work with WhatsApp?
Yes, WhatsApp is the main channel. Because it runs on top of Chatwoot, it also handles Instagram, website chat, email, and other channels, all in one place.
Does my data stay within my company?
Yes. It is self-hosted: it runs on your own infrastructure, with your data and conversations under your control.
Can I use local AI models?
You can. Beyond the top commercial models, the agent works with local models through a compatible endpoint, if you'd rather not depend on an external cloud.
The agents run on top of Chatwoot. Discover Chatwoot fazer.ai.
Your WhatsApp answering today.
Paste the command, let the assistant set everything up, and start answering.
$ curl -fsSL https://app.fazer.ai/api/agents/install.sh | bash