Em português: exemplos prontos de uso do Turbo Notify, o WhatsApp para o seu sistema, os seus agentes de IA e os seus clientes. Cada exemplo é um projeto independente: clone, preencha as variáveis de ambiente e rode. O primeiro,
agno-support-agent, é um atendente de WhatsApp feito com Agno que responde pelo servidor MCP do Turbo Notify. Para montar o seu próprio agente passo a passo, veja o guia de agentes de IA. O restante deste repositório está em inglês.
Working examples of building on Turbo Notify: WhatsApp for your software, your AI agents and your customers.
Each example is a self-contained project: clone it, set a few environment variables, and run it. Nothing here depends on anything else in this repository, and nothing here needs access to Turbo Notify's internals. Every example talks to the same public API and the same MCP server your own code would.
| Example | What it shows |
|---|---|
| agno-support-agent | A WhatsApp support attendant that answers questions about Turbo Notify, built with Agno and driven entirely through the Turbo Notify MCP Server. Receives inbound messages by webhook, has an LLM write the answer (its only tool reads one message), and sends it as a normal message through the MCP server. |
More are coming. If you build something worth showing, open a pull request.
- A Turbo Notify account with at least one connected WhatsApp number (in the dashboard, under Numbers, or Números when the dashboard is in Portuguese).
- An API key from the dashboard, under API Keys.
- Whatever the individual example lists in its own README.
They do not assume you have read the whole API reference first. That is the point: each one is meant to be readable start to finish in a few minutes.
Several examples reach Turbo Notify through the Model Context Protocol rather than by calling REST endpoints directly. That is not a different API: it is the same public API, presented as tools an LLM can choose between, with schemas it can read.
If you are building an agent, that is usually what you want: you describe the goal, and the model
picks get_message or get_number on its own. Be sparing with the tools that send: the safest
agent is one whose model has no send tool at all, writes the answer as text, and leaves the sending
(send_text, to the sender the webhook named) to your own code, as agno-support-agent does. If
you are building a deterministic integration, call the REST API directly instead; both are
supported and both are documented.
Building an agent of your own? Start with the AI agent guide in the documentation: it walks through the same pattern these examples use (webhook in, MCP out) and how to keep a message from a stranger from steering your agent.
MIT. Copy anything here into your own project.