WhatsApp MCP: How to Connect WhatsApp to AI in 2026
No — there’s no official WhatsApp MCP server from Meta. What you’ll find instead is a handful of community-built servers with real adoption (the most-starred is lharries/whatsapp-mcp, plus business-focused forks that wrap Meta’s own Cloud API), but none of them are published or maintained by Meta itself. If you’re searching “WhatsApp MCP” hoping for an official connector like HubSpot or Notion ship, it doesn’t exist yet.
That gap matters less than it sounds, because even an official MCP server would only hand WhatsApp to an AI inside a conversation you start. It’s a doorway, not a worker — nothing watches your chats for you, nothing fires when a message arrives, and nothing runs once you close the window. Here’s what the community WhatsApp MCP servers actually do, the honest way to set one up, where the whole approach stops — and what to use when you want WhatsApp work that runs on its own.
What an MCP connection to WhatsApp does
Model Context Protocol (MCP) is the open standard that lets an AI client — Claude, ChatGPT, Cursor, and others — talk to an outside app through a shared interface. Since Meta hasn’t shipped one, two different flavors of community server have filled the gap:
- Personal WhatsApp Web servers (like
lharries/whatsapp-mcp) — a bridge that logs into your personal WhatsApp account via QR code, using the same reverse-engineered WhatsApp Web protocol (whatsmeowor Baileys) that unofficial WhatsApp tools have used for years. It stores your message history locally and exposes it to an MCP client. - Business Cloud API servers (like
tkhattar14/whatsapp-business-mcp) — a thin MCP layer over Meta’s own, official WhatsApp Business Platform (Cloud API), so calls actually hitgraph.facebook.com. This is the more defensible option if you’re messaging customers rather than personal contacts.
With either connected, an AI client can:
- Search and read message history — “find the last thing Jamie said about the invoice” answered from real chat data.
- Look up contacts and threads — surface a conversation without scrolling for it.
- Draft and send messages — reply to an individual or group from inside the chat.
- Summarize a backlog — turn a day of unread messages into a short list of what needs a reply.
It’s useful for ad-hoc lookups. It is not a Meta product, and the personal-account flavor means running an unofficial client against WhatsApp’s protocol — outside WhatsApp’s terms of service and carrying some risk of account action.
How to connect WhatsApp to AI
There’s no vendor install page to point you to, because there’s no vendor server. The realistic paths:
- Community personal server — clone
lharries/whatsapp-mcp(or a maintained fork), run its local bridge, and scan a QR code to link it to your WhatsApp account the same way you’d link WhatsApp Web. Add the resulting MCP server to your AI client’s config. - Business Cloud API server — register a WhatsApp Business Platform app in Meta’s developer console, generate an access token, then point a community MCP server (or your own thin wrapper) at the official Graph API endpoints using that token.
- Self-host and maintain both — either path means you’re running a server, storing credentials, and keeping the bridge updated yourself. Nobody official is patching it for you.
Where a WhatsApp MCP setup stops
Even setting the “unofficial” issue aside, the same four limits that cap every MCP integration apply here:
- It only works inside a chat you start. Close the window and nothing happens. No server watches your WhatsApp threads; it waits for you to ask.
- No triggers. A new message from a client, a group hitting a certain size, a keyword appearing in a thread — none of these can start anything through MCP. There’s no “when this happens in WhatsApp, do that.”
- It’s one app at a time. The WhatsApp MCP knows WhatsApp. Getting a customer’s message into your CRM, a task list, and a Slack channel means wiring up a separate MCP server for each and hoping your client can juggle all of them in one turn.
- You own the plumbing, the scopes, and the risk. Credentials, token refresh, and — for the personal-account flavor — the standing risk to your WhatsApp account are all on you, not on a vendor’s support team.
So a community WhatsApp MCP server is a way to ask about your messages and send an occasional reply from a chat. It is not a way to make WhatsApp run — to have work happen on a schedule or in reaction to an incoming message, across the other tools that conversation touches.
Running WhatsApp work that doesn’t need a chat open
That “run on its own, across apps” gap is exactly where Carly fits. Carly connects to WhatsApp natively — no QR-linked bridge to babysit, no Graph API token to rotate — and to the ~260 other apps it supports, plus anything with a public API through your own key. The difference from MCP is the important part: Carly’s workflows are triggered and scheduled, so WhatsApp work happens whether or not anyone has a chat window open.
A few things that MCP can’t do but a Carly workflow can:
- When a message comes in from a client → log it to your CRM, create a follow-up task, and draft a reply for approval — automatically, the moment it arrives.
- Every evening → summarize the day’s unread WhatsApp threads and email the list to you before you log off.
- When a keyword like “invoice” or “urgent” appears → pull the sender’s contact record, flag the thread, and post a heads-up to the right Slack channel.
The non-AI steps — the moving, matching, and routing between apps — are free and unlimited, the Zapier-style backbone of the workflow. The AI steps (drafting, summarizing, deciding) start at $35/month. You describe the outcome in plain language and Carly wires up the WhatsApp connection and everything downstream.
If you just want to ask an AI about your WhatsApp messages from a chat, one of the community servers will technically do it — with the caveats above. If you want WhatsApp to actually do things — on a trigger, on a schedule, across every app a conversation touches — that’s the job MCP wasn’t built for, and it’s the one Carly was.
FAQ
Does WhatsApp have an official MCP server?
No. Meta hasn’t published one. What exists are community-built servers — some (like lharries/whatsapp-mcp) connecting to personal accounts through the unofficial WhatsApp Web protocol, others wrapping Meta’s official Business Cloud API for business messaging.
Is it safe to use an unofficial WhatsApp MCP server on my personal account? It carries some risk. Personal-account servers rely on reverse-engineered libraries that sit outside WhatsApp’s terms of service, which means there’s a chance of account action from Meta. The Business Cloud API route, since it uses Meta’s own official API, is the more defensible option for business use.
Can a WhatsApp MCP server trigger automations? No. MCP is request/response inside an AI chat — it has no triggers and nothing runs when the conversation is closed. For event- or schedule-driven WhatsApp work across apps, you need a workflow tool like Carly rather than an MCP server.
What AI tools can connect to WhatsApp over MCP? Any MCP-compatible client — Claude, ChatGPT, Cursor, and others — can connect to a community WhatsApp MCP server once you’ve set one up and pointed the client at it.
Can I connect WhatsApp to AI without coding or hosting a server? Yes. You don’t have to touch MCP, QR-code bridges, or Graph API tokens at all. Carly connects to WhatsApp for you and lets you build the automation in plain language — describe what you want to happen and it wires up WhatsApp and the other apps involved, with no server to host and no code to write.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
