Claude Cowork and WhatsApp: The Only Route That Works
There is no WhatsApp connector for Claude, so there is none for Cowork either. Anthropic’s directory has entries for Slack, for Twilio, for a few hundred other apps. claude.com/connectors/whatsapp returns a 404. The only route in is a custom MCP server you run yourself against the WhatsApp Business API.
That is doable, and worth knowing precisely, because the near-miss on the way is a trap.
The Twilio listing is not the shortcut it looks like
Twilio is in the directory, it is categorised under Communication, and Twilio is how a great many people send WhatsApp messages. So it looks like the answer.
Read the capability badge. Twilio’s is Read, and the description explains why: it “surfaces the exact specs and code patterns you need” across more than 1,800 endpoints. Its own worked example is the giveaway.
“What is the endpoint for sending a WhatsApp message?”
That is a documentation search tool. It tells Cowork how to send a WhatsApp message. It does not send one. Adding the Twilio connector and expecting messages to go out is the single most common wrong turn here.
What the working setup looks like
- Get access to the WhatsApp Business API, either directly through Meta or through a provider. Personal WhatsApp has no supported API, so a consumer account is not a starting point.
- Run a third-party WhatsApp MCP server configured against that account. Capabilities vary by server, and a Business-API-backed one typically reads threads and sends messages the account has access to.
- In Claude, open Customize → Connectors, click the + next to Connectors, choose Add custom connector, and enter the server’s name and URL, plus OAuth client ID and secret if it needs them.
Two constraints from Anthropic’s own documentation shape this. Custom remote MCP connectors are available “on Claude, Cowork, and Claude Desktop” for free through Enterprise plans, with free accounts limited to a single one, so the connector itself is not what costs money. And custom connectors are called from Anthropic’s cloud rather than from your machine, which means the server has to be reachable on the public internet. A localhost MCP server will not work here.
Once connected, set the tool permissions deliberately. Under Tool permissions for that connector, message-sending tools should sit at Needs approval rather than Always allow until you trust it, and Cowork’s own mode selector should stay on Manually approve for the first few runs. Outbound messages to real customers are the textbook case for staying close.
What you get, and the wall you hit
With a server running, a Cowork task can read the account’s threads, pull context out of a conversation and draft or send a reply as part of a longer job. That composes well with the rest of Cowork: read the thread, check the calendar, build the quote, send the message.
Then you hit the wall that has nothing to do with WhatsApp.
MCP has no trigger mechanism and no scheduler. Its own Triggers and Events working group opens by conceding that clients learn about server-side changes by polling, and that work is still listed as on the horizon rather than shipped.
Cowork’s /schedule command runs a task daily, weekly or monthly in Anthropic’s cloud with no device online, and that is genuine unattended work. It is still a clock. A scheduled task can check WhatsApp every hour and act on what it finds. Nothing fires the moment a customer messages you, which for a messaging channel is most of the point. Customers who message on WhatsApp expect a reply in minutes, and an hourly poll is a poor imitation of that.
WhatsApp that answers on arrival
Carly connects WhatsApp natively, alongside Slack, Telegram and Discord, from its 260+ integrations, so there is no server to host and nothing to expose to the public internet. You connect the account and it is on.
The difference that matters is the trigger. An inbound WhatsApp message starts the work: Carly reads the thread, checks your calendar, answers the question, books the slot, files the record, and hands you the ones that need a human. The same agent covers the email side of the same customer, because it holds several mailboxes and channels at once rather than one identity per connector.
Zapier-style workflow steps are free, so the routing and logging cost nothing to run, and AI agents start at $35 a month for the parts that need judgment. Set the first one up at carlyassistant.com. For high-volume or embedded messaging, the API path is available too, and the fastest way in is to book a call.
FAQ
Is there a WhatsApp connector for Claude Cowork?
No. WhatsApp is not in Anthropic’s connectors directory, so the only route is a custom MCP server built on the WhatsApp Business API and added under Customize → Connectors.
Can I use the Twilio connector to send WhatsApp messages from Cowork?
No. The Twilio listing’s capability is Read, and it searches Twilio’s API documentation rather than calling those APIs. It will describe the endpoint for sending a WhatsApp message without sending one.
Does connecting WhatsApp to Cowork need a paid Claude plan?
Custom MCP connectors work on Free as well as paid plans, with Free limited to one. Cowork itself is what requires a paid plan, starting at Pro.
Can Cowork reply to a WhatsApp message the moment it arrives?
No. Cowork tasks start when you start them or when a schedule fires, so the fastest unattended pattern is a scheduled poll rather than a reply on arrival.
Does this work with personal WhatsApp?
Not in a supported way. The MCP servers in this space are built on the WhatsApp Business API, which a personal account does not expose.
More: Claude Cowork connectors · Claude WhatsApp integration · WhatsApp MCP · Claude Cowork limitations · Claude Cowork use cases
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."
