Skip to content

Glossary

WhatsApp Business and AI agents glossary

30 terms explained in plain language: from the 24-hour window and message templates to MCP, RAG and the names Meteor uses.

What is in this glossary?

Short definitions of the 30 terms that come up most when a business in Latin America automates WhatsApp with AI agents: Meta’s rules for the WhatsApp Business API, the parts of an AI agent and the product names in Meteor. Each term starts with a direct definition, followed by what it means in practice and a link to the page that covers it in depth.

The WhatsApp terms follow Meta’s official documentation, linked under each definition with the date it was consulted.

Last updated

WhatsApp Business

Terms from the Meta platform: accounts, sending rules and billing.

WhatsApp Business API (WABA)

The WhatsApp Business API is Meta’s official way for a business to send and receive WhatsApp messages from software, at scale. WABA stands for WhatsApp Business Account, which groups the company’s phone numbers; templates and billing live in its messaging account. It is the foundation for any serious WhatsApp operation.

It is what lets you connect a number to an AI agent, a shared inbox or your CRM. The account belongs to your company in its Meta business portfolio, even when a provider connects it. Meteor connects your line through this official API: the account stays yours.

Read more: What the WhatsApp Business API is and how to get it

Official source: Meta: About the WhatsApp Business Platform , consulted .

WhatsApp Business app vs. the API

The WhatsApp Business app is the mobile app built for small businesses: one person answers one conversation at a time from a phone. The API is the connection for software: several people and AI agents serve the same number, with templates, integrations and sending at scale. Both use the same WhatsApp network.

The app works while one or two people can keep up. When messages outgrow the team, or you want an agent to answer with data from your store or ERP, you need the API. With coexistence you do not have to choose: the same number can stay on both.

Read more: WhatsApp API or the WhatsApp Business app: which one fits you

Official source: WhatsApp Business: the WhatsApp Business app (Spanish) , consulted .

Coexistence

Coexistence is Meta’s option to connect a number that already uses the WhatsApp Business app to the API, without leaving the app. The number works on both sides at the same time, and when you connect it you can sync contacts and up to 180 days of earlier conversations.

It answers a small business’s most common objection: “I don’t want to lose my WhatsApp.” In Meteor your contacts and up to six months of conversations are imported; from that history, files older than 14 days do not come through. It depends on Meta enabling coexistence for your number.

Read more: WhatsApp AI agent

Official source: Meta: Onboarding WhatsApp Business app users (coexistence) , consulted .

BSP or WhatsApp solution provider

BSP (Business Solution Provider) is the common name for a company that connects other businesses to the WhatsApp Business API and offers software on top. Meta organizes those partners as Solution Partners, Tech Providers and Tech Partners: Solution Partners can extend credit and invoice API usage; the others cannot.

With a Tech Provider, the client adds its own payment method and Meta bills it directly for API usage. Before you sign, ask who invoices Meta’s messages, whether there is a markup and whose name the account is in. In Meteor what Meta charges is separate from the plan, with no markup, and the account stays yours. An unofficial API goes through no authorized provider and puts the number at risk.

Read more: Unofficial WhatsApp API: what you risk

Official source: Meta: Partners , consulted .

Official Business Account (blue badge)

An Official Business Account is a WhatsApp Business number whose company Meta has verified as authentic. You recognize it by the blue badge next to the name. To request it, the company must pass business verification, have been on the platform for 30 days, enable two-step verification and have an approved display name.

The badge builds trust, but it is not required to use the API or to connect an AI agent. Meta does not grant it to numbers that are only on the app. If it is denied, you can request it again after 30 days.

Read more: What the WhatsApp Business API is and how to get it

Official source: Meta: Official Business Accounts , consulted .

24-hour customer service window

The 24-hour customer service window is the period that opens when a customer messages or calls you on WhatsApp. While it is open, you can reply with free-form messages; every new customer message restarts it. Once it closes, you can only write to that customer with a template approved by Meta.

It is the rule that causes the most confusion when automating: an AI agent replies freely inside the window, but a follow-up the next day needs a template. In Meteor, the Met writes outside the window only with an approved template. On Instagram and Messenger, after 24 hours you cannot write at all.

Read more: How much the WhatsApp Business API costs

Official source: Meta: Pricing on the WhatsApp Business Platform , consulted .

Message template

A message template is a predefined text, with variables, that Meta reviews before you can send it. It is the only type of message that can be sent outside the customer service window. Every template belongs to one category: marketing (promotions), utility (updates about a transaction) or authentication (verification codes).

Review can take up to 24 hours, and approval is not permanent: if customers block or report it, Meta can pause it. The category sets the rate, so do not dress a promotion up as an update. In Meteor, Broadcasts and reminders outside the window use approved templates.

Read more: How much the WhatsApp Business API costs

Official source: Meta: Message templates overview , consulted .

WhatsApp per-message pricing

Per-message pricing is how Meta has charged for the WhatsApp Business API since July 1, 2025: it charges each delivered message, not each conversation. The price depends on the message category and on the country of the receiving number. Messages that customers send you are not charged.

Since October 1, 2026 Meta also charges for service messages, the replies inside the window, with 1,000 free per number each month. To estimate your cost, separate how many marketing, utility and authentication templates you send and to which countries. In Meteor that charge is Meta’s and is separate from the plan.

Read more: How much the WhatsApp Business API costs

Official source: Meta: Pricing on the WhatsApp Business Platform , consulted .

WhatsApp opt-in

Opt-in is the permission a person gives a business to message them on WhatsApp. Meta requires it before the first message: it must clearly say that the person agrees to receive messages and name the business, and it must comply with local law. You choose how to collect it.

A web form, a text message, an in-store signup or a phone call with an automated menu all work, as long as it is clear. In Latin America each country’s data protection law applies too, such as habeas data in Colombia. Asking per message type and explaining how to opt out protects your number’s quality.

Read more: Habeas data and AI agents on WhatsApp

Official source: Meta: Getting opt-in for WhatsApp , consulted .

Quality and messaging limits

The messaging limit is how many different customers your business can contact outside the customer service window in 24 hours. It starts at 250. It moves to 2,000 once you verify the business or send 2,000 messages in 30 days with high-quality templates; then it scales automatically to 10,000, 100,000 or unlimited.

To scale automatically, Meta requires high-quality messages across all your numbers and templates, and that you used at least half of your limit in the last 7 days. The limit belongs to the business portfolio and all its numbers share it. Meteor measures each line’s daily sends against its tier: it warns at 80 % and stops sending at 100 %.

Read more: WhatsApp AI agent

Official source: Meta: Messaging limits , consulted .

WhatsApp catalog

The WhatsApp catalog is the inventory of products or services a business uploads to Meta and connects to its number, with name, description, price, images and availability. The customer browses it inside the chat, builds a cart and sends the order without leaving the conversation. Businesses receive those orders as messages.

It is loaded from Meta Commerce Manager or by API, and Meta reviews every item. If your inventory changes daily, keeping two catalogs up to date is costly. A Meteor Met answers with your Collections or with the catalog of your connected Shopify or WooCommerce store, and checks real stock.

Read more: AI agents for e-commerce

Official source: Meta: Sell products and services , consulted .

AI agents and automation

How an AI agent works, what connects it to your systems and where it can fail.

AI agent

An AI agent is software that receives a goal, decides the steps and uses tools with permissions to achieve it: it looks up data, takes actions and records what it did. Unlike a chatbot that only talks, an agent can finish a task, with or without a conversation in between.

For a business, the practical difference is which data it looks up and which action it takes. A useful agent confirms a payment, books an appointment or moves a deal in the CRM, and hands off whatever needs judgment. Always ask for its list of tools and the permission each one uses.

Read more: AI agents vs. chatbots

WhatsApp AI agent

A WhatsApp AI agent handles the conversations on a WhatsApp Business line: it keeps each contact’s context, answers with your business information and uses the tools you authorized to give real data or take the next step. Whatever needs human judgment, it hands to your team with the context.

In Latin America WhatsApp is where questions about price, stock, appointments and orders arrive, at any hour. An agent connects through the official API and works with messages, voice notes and files. Outside the 24-hour window, it only writes with an approved template.

Read more: WhatsApp AI agent

Flow or automation

A flow is a defined sequence of steps, conditions and waits that runs the same way every time an event happens or a rule is met. An automation is that flow running without anyone triggering it by hand. It can include an AI agent in one step without the whole path being autonomous.

It pays to split the work: the agent reasons about what changes from case to case, and the flow always repeats what cannot fail, like charging, shipping or notifying. In Meteor flows are designed on a visual canvas with more than 50 node types, with the Met inside.

Read more: Step-by-step automations

Conversational CRM

A conversational CRM is a CRM fed by conversations: each chat is tied to its contact, its owner, its pipeline stage, its notes and its tasks. The value is not in storing messages but in turning them into a record the team can pick up and measure. It closes the gap between talking and selling.

Without it, follow-up depends on searching and copying chats by hand. With an AI agent connected, the contact is recognized or created automatically, context lands in custom fields and the stage moves while the customer is served. Ask to see the record a real conversation leaves behind.

Read more: AI CRM for WhatsApp (in Spanish)

Shared inbox

A shared inbox is the view where several people on a team handle the conversations of one or more numbers and channels from one place. Each chat has an owner, a status and internal notes, and the team sees who is handling it, so nobody answers twice and no customer is left waiting.

It is what the WhatsApp Business app does not give you once the team grows. In Meteor, if you have several WhatsApp numbers, each line is its own inbox with its own filters, and round-robin assignment among the people in a group ships turned off and is enabled per group.

Read more: AI CRM for WhatsApp (in Spanish)

Human handoff

Human handoff is the moment an AI agent passes a conversation to a person on the team, because information is missing, there is an exception, the customer asks for it or the decision needs judgment. A good handoff also passes the context: who the customer is, what they asked for and what is still pending.

A “talk to an agent” button is not enough if the person has to reread the whole chat. In Meteor a transfer is a single action: it leaves an internal note with the summary and the reason, pauses the Met in that chat and assigns the person.

Read more: AI CRM for WhatsApp (in Spanish)

RAG and knowledge base

RAG (retrieval-augmented generation) is the technique by which an AI agent searches for information in a source, such as a knowledge base, and uses it to answer. The knowledge base holds the content that can be retrieved: frequently asked questions, policies, manuals or product sheets. It grounds answers in your own content.

RAG is for what is written down, not for what changes every minute: stock or an order’s status is looked up in the system that holds it. In Meteor, Collections organize knowledge and connectors look up live systems.

Read more: Company Brain: the agent that knows your company

MCP (Model Context Protocol)

MCP (Model Context Protocol) is an open standard for connecting AI applications to external systems: data sources, tools and workflows. An MCP server exposes those capabilities in a common format, and any compatible client, such as Claude, ChatGPT or Cursor, can use them without a custom integration.

MCP standardizes the connection, but it does not decide permissions: what the agent can read or change is still a decision for whoever configures it. Meteor exposes a remote MCP server to operate your account from Claude, Cursor and other clients, limited to your key’s scopes.

Read more: Meteor API, SDK and MCP

Official source: Model Context Protocol: What is MCP? , consulted .

Connector or integration

A connector is the piece that links a platform with another system, such as a store, an ERP, a CRM or a payment gateway, through its API. It defines which data can be read and which actions can be taken. Having a connector does not mean having access to the entire external system.

The scope depends on the credential and on the operations the connector implements. Before deciding, check the list of actions and what is left out. In Meteor, every integration page says what the connector does and what it does not.

Read more: Meteor integrations

Webhook

A webhook is an automatic notice that one system sends to another over the internet when something happens, such as an approved payment, a new order or a received message. Instead of asking every few minutes whether there is anything new, the receiving system reacts at that moment.

It is the cheapest and fastest way to connect two systems: it spends no requests when nothing happens. WhatsApp delivers messages and delivery statuses to your platform this way. In Meteor, a flow can start when a webhook arrives, and the developer portal documents webhooks and events.

Read more: Meteor API, SDK and MCP

Public API and SDK

A public API is the documented door through which other software uses a platform: it reads data and takes actions with a key and a set of permissions. An SDK is a library for a programming language that wraps that API, so you call it with functions instead of building each request by hand.

It is useful when your technical team wants to bring the platform into its own product or automate from its code. In Meteor keys are issued with granular scopes and test keys, and there are TypeScript and Python SDKs and a CLI.

Read more: Meteor API, SDK and MCP

Function calling (tools)

Function calling is the mechanism by which an AI model asks to run a tool: a function with a name, defined inputs and a result, such as finding a contact, checking stock or creating an appointment. The model decides when to call it; the system runs it and returns the result to the model.

It is what separates an agent that talks from one that acts. “Integrates with the CRM” says nothing; the list of tools, and which ones read and which ones write, does. In Meteor every integration page lists the connector’s actions.

Read more: Meteor CRM integration page, tool by tool

Hallucination

A hallucination is an answer from an AI model that sounds right but is not backed by the available data: an invented price, a policy that does not exist or an order that was never placed. It happens because the model produces probable text, not verified text, and it can sound fully confident.

Connected sources, tools and clear instructions reduce it, but no technique removes it. A responsible design defines what the agent does when it lacks the data: say so, ask for information or hand the case to a person, instead of guessing.

Read more: Glossary of chatbots and AI agents

Prompt or agent instructions

An agent’s instructions, or system prompt, are the text that tells it who it is, what to do, in which tone, what it cannot promise and when to hand the case to a person. It reads them before every answer, which is why they are the most direct lever for changing its behavior.

Good instructions are concrete: which data to ask for, what to answer when information is missing and which cases to escalate. They do not replace tools: if the agent cannot check stock, no instruction gives it the number. In Meteor each Met’s instructions are adjusted from the platform.

Read more: How to create a WhatsApp AI agent

Lead qualification

Qualifying a lead means deciding whether a prospect has the need, budget, timeline and decision power to buy, and how urgently the sales team should handle them. The result is a stage or a tag in the CRM that says what to do next with that contact, and who should do it.

An AI agent asks the questions while chatting and stores the answers in the CRM, but your company defines what counts as qualified. Write it down before automating: if the team cannot agree, the agent cannot either.

Read more: What it means for a lead to be qualified

In Meteor

The product’s own names, as they appear in the platform.

Met

A Met is a Meteor AI agent. Conversational Mets serve, qualify, quote or book on WhatsApp, Instagram, Messenger and web chat; operational Mets work on connected systems even when the task does not go through a conversation. Each Met has its own instructions, tools and limits.

No plan limits how many Mets you create: what is charged is the work they do, as Energy. That way you can split into two Mets a job one used to do, such as service and sales. The name identifies the product; it is not a general industry category.

Read more: Meteor’s Mets

Energy

Energy is how Meteor charges for AI usage: every job a Met does consumes Energy according to its complexity. Every plan includes an Energy bonus in each period you pay for, from US$10 to US$450 depending on the plan, and you can top up whenever you need to.

It separates the plan price, which covers team, contacts and channels, from actual AI usage. In a Room, a message between people uses no Energy; a Met at work does. What Meta charges for WhatsApp messages is separate and is not Energy.

Read more: Meteor pricing

Room

A Room is a Meteor work chat where several people in your business and their Mets talk together, up to 30 members per room. People talk among themselves, call a Met with @ when they need it to do something, and what the Met produces stays in the room for someone to approve.

It lets the team decide in one place what gets published: a carousel, an ad or a piece of copy. A message between people uses no Energy. Your customers do not enter rooms: their conversations stay in the inbox.

Read more: Meteor Rooms

Routine

A Routine is a complete business process that the team of Mets learns and repeats: fixed steps, one Met in charge of each step, and any missing setup shows up along the way. Whatever publishes, charges or spends waits for a person’s approval before it goes out.

An automation is a task; a routine is the process that contains it, such as serving and selling, posting on social media or taking bookings. Each routine links to the automations it is made of.

Read more: Meteor Routines

Looking for the difference between a chatbot and an AI agent, memory and context, or human approval? The glossary of chatbots and AI agents covers them with a test you can run in any demo.

See these terms working on your own WhatsApp

In a demo we connect the dots with your case: your line, your templates, your systems and what your Mets should do.

I want a demo