Introduction
Este conteúdo não está disponível em sua língua ainda.
Wabery gives agencies, businesses, and developers one project for customer conversations, a hosted agent, website chat, WhatsApp, lead qualification, and reservations. Use the dashboard without writing code or build through the API when you need your own integration.
Build through the website, an AI coding agent, the REST API, or the CLI. They all work on the same project.
Get startedWho it is for
Section titled “Who it is for”- Agencies shipping a WhatsApp chatbot per client, isolated in its own project
- Developers who want a stable API instead of wiring Meta Cloud API by hand
- AI coding agents that can read these docs and call Wabery through MCP
Wabery is not for unofficial WhatsApp Web automation, personal accounts, or unsolicited messaging.
First decision: who writes the replies?
Section titled “First decision: who writes the replies?”Every project has one primary reply mode:
| Mode | Pick it when | What happens |
|---|---|---|
| Wabery AI agent | You need Wabery to provide the chatbot | Choose a customer outcome, add knowledge and optional connectors, test the Draft privately, and publish it when the required channels and modules are ready. |
| Your own backend | You already have an agent, app, n8n workflow, Make scenario, or Zapier automation | Wabery sends inbound messages to your webhook. Your system decides what to say and sends the reply through the API. |
This is separate from using Claude, ChatGPT, Cursor, or Codex through MCP. Those are developer tools: they help you configure the project, but they do not sit online waiting for customer messages.
You choose the reply mode when you create a project. You can change its routing later from the Agent workspace after the replacement path is ready.
What you can do
Section titled “What you can do”| You want to | Start here |
|---|---|
| Have Wabery run the chatbot | Hosted agent |
| Connect your existing backend or workflow | External agent |
| Test either path with no Meta setup | Sandbox quickstart |
| Manage confirmed tables, classes, or appointments | Reservations |
| Collect confirmed customer enquiries | Lead qualification |
| Add chat to an approved website | Website chat |
| Share a limited workspace with a client | Client project access |
| Send and receive messages | Sending messages |
| Collect structured answers in chat | WhatsApp Flows |
| Broadcast an approved template | Broadcasts |
| Reply from your own server | Webhooks |
| Let Claude, ChatGPT, or Cursor build | MCP server |
How a project is organized
Section titled “How a project is organized”A project is one client, business, or product. It can have:
- Channels — the WhatsApp Sandbox, a business number, website chat, or a supported social channel
- Inbox — conversations you can read, reply to, and hand off to a person
- Agent — either Wabery’s hosted agent or your external integration
- Bookings and Leads — optional workflows controlled separately from agent publication
- Project access — View-only or Editor access for the client, without organization billing or credentials
- Developer access — API keys and connections for ChatGPT, Claude, Codex, the CLI, or your application
You do not need every piece on day one. Choose who replies, configure that path, then test it in the sandbox.
Ways to build
Section titled “Ways to build”Dashboard. Work from Inbox, Agent, Bookings, Leads, Channels, Settings, and the Developer section. Project tools appear only when the project can use them.
API and SDK. https://api.wabery.com/v1 with a project-scoped Bearer key.
The machine-readable contract is the OpenAPI spec.
CLI. npm i -g @wabery/cli, then wabery login. Same project, from a
terminal.
MCP. Point Claude, ChatGPT, Cursor, or Codex at https://mcp.wabery.com.
It can configure Wabery for you. It is a build tool, not the customer-facing
runtime.