WhatsApp MCP server: what it is and the options in 2026
What a WhatsApp MCP server does for AI assistants, the options as of November 2026 (Meta's beta server, inbox tools, n8n) and where webhooks fit in.
- Written by
- By WavesFlo team
- Published
- Published
- Reading time
- 5 min read
A WhatsApp MCP server lets an AI assistant such as Claude or ChatGPT read and act on WhatsApp through tools, using the Model Context Protocol. As of November 2026 there are three kinds: Meta's own server for its Cloud API, servers from inbox tools over the numbers they connect, and servers you build yourself, for example in n8n.
This guide explains what an MCP server is, what each option reaches, what to check before giving an assistant a live WhatsApp number, and how signed webhooks fit beside it.
What is an MCP server?
The Model Context Protocol is a standard way to connect an AI assistant to outside systems. An MCP server publishes three kinds of thing: tools, functions the model can call to take an action; resources, file-like data it can read for context; and prompts, predefined instructions for a task. Each tool comes with its name, description, inputs and outputs, so the assistant knows when and how to call it.
For WhatsApp, the tools are the everyday jobs: list chats, search messages, send a reply, label or assign a chat. The assistant decides which to call from what you ask it.
What WhatsApp MCP servers are there?
| Option | What it reaches | What its tools do | Worth knowing |
|---|---|---|---|
| Meta's WhatsApp Business Tools MCP | WhatsApp Business Platform (Cloud API) accounts | 18 tools: find businesses, accounts and numbers; add and register numbers; manage pre-approved formats; set up webhooks; send messages | In beta and rolling out gradually; sign in with a Meta developer account; rate limited per user and per tool |
| TimelinesAI's MCP server | The WhatsApp numbers connected to TimelinesAI | List chats, search messages, send replies, add labels, assign teammates, check quotas | On all its plans; the assistant sees only what your TimelinesAI role allows; every call counts against its monthly API allowance |
| Periskope's MCP server | The WhatsApp numbers connected to Periskope | Chats, groups, contacts and tickets, with every action its API offers | Hosted, with sign-in by consent screen; its page does not say which plans include it |
| Your own, built in n8n | Whatever your workflows call | n8n's MCP Server Trigger node publishes tool nodes and workflows as tools | Works over server-sent events or streamable HTTP, not stdio; optional bearer, header or n8n sign-in |
Which one fits follows from where your numbers already are. A Cloud API number points to Meta's server. Numbers already connected to an inbox tool point to that tool's server, if it has one. Anything else, including your own systems around WhatsApp, can be reached through a server you build.
Is a WhatsApp MCP server safe to use?
It is as safe as the permissions you grant. Meta's own documentation warns that an AI agent can call anything your granted permissions allow, without asking you first. On a live number, a send reaches a real customer. Before connecting one:
- Grant read-only access first, and add sending only when you have watched what the assistant does with it.
- Use an account or role that sees only the numbers and chats the assistant needs.
- Keep WhatsApp's own rules in view: message people who asked to hear from you, and stop when they say stop.
- Read the server's limits: API call allowances, rate limits and message quotas still apply when an assistant is the caller.
How do webhooks fit beside an MCP server?
They are the two directions of one connection. An MCP server is the pull side: the assistant calls a tool when someone asks it to. A webhook is the push side: the system calls your URL the moment something happens, such as a message arriving, someone joining a group or a number disconnecting.
WavesFlo's developer reference publishes a REST API and signed webhook events across messages, groups, consent, campaigns and numbers. n8n connects to WavesFlo through the REST API and signed webhooks, so one n8n instance can hold both halves:
- A Webhook node receives WavesFlo's signed events and checks each signature, so the workflow starts the moment a message or group change arrives.
- Workflows that call WavesFlo's REST API, such as sending a reply or reading a conversation, do the work.
- An MCP Server Trigger node publishes those workflows as tools, so an assistant can call them with the authentication you set.
Sends through the API meet the same opt-outs, daily caps, cooldowns and quiet hours as a campaign from the dashboard, whoever or whatever made the call. WhatsApp automation with webhooks and n8n covers signatures and retries, and n8n to WhatsApp groups has workflows you can copy.
Questions
What is a WhatsApp MCP server?
A server that publishes WhatsApp actions, such as listing chats, searching messages or sending a reply, as tools an AI assistant can call through the Model Context Protocol.
Does Meta have its own WhatsApp MCP server?
Yes. Meta's WhatsApp Business Tools MCP works with WhatsApp Business Platform (Cloud API) accounts. As of November 2026 it is in beta and rolling out gradually, so it may not be available to every account yet.
Do I need the Cloud API to use an MCP server with WhatsApp?
No. Meta's server needs a Cloud API account, but inbox tools such as TimelinesAI and Periskope publish servers over the numbers connected to them, and you can build your own in n8n.
Can an AI assistant post into WhatsApp groups through MCP?
It depends on the server. Periskope's lists group actions; Meta's reaches Cloud API numbers, where groups go through its Groups API; a server you build can do whatever the workflows behind it do.
How do I build a WhatsApp MCP server with n8n?
Add n8n's MCP Server Trigger node, attach the tool nodes or workflows you want to publish, set authentication, and give your assistant the trigger's URL. The workflows themselves call your WhatsApp tool's API.
Sources
- Microsoft Learn: Extend your agent with Model Context Protocol (checked 7 Nov 2026)
- Meta for Developers: MCP servers (checked 7 Nov 2026)
- Meta for Developers: WhatsApp Business Tools MCP (checked 7 Nov 2026)
- TimelinesAI Help: Connect TimelinesAI to your AI assistant (MCP), updated 6 Oct 2026 (checked 7 Nov 2026)
- Periskope: WhatsApp MCP (checked 7 Nov 2026)
- n8n Docs: MCP Server Trigger node (checked 7 Nov 2026)