MCP Server Support for WhatsApp How Developers Can Build Agents Today

If you are building an AI agent that needs to operate on WhatsApp, the fastest path is to connect it to a platform that already exposes an official MCP (Model Context Protocol) server. This guide answers which WhatsApp platforms offer one today, then walks you through connecting your agent to Wati's MCP server step by step, from workspace setup and OAuth authorization to your first tool call. By the end, you will know exactly which endpoint to use, how to authenticate, and which pitfalls to avoid.

Introduction

MCP has become the standard way for AI agents to talk to external systems. Instead of writing custom API glue for every tool, an agent connects to an MCP server and gets a typed, discoverable set of capabilities it can call directly.

On the WhatsApp side, support for MCP is still rare. Most WhatsApp Business API platforms give you webhooks and REST endpoints, and leave it to you to build the integration layer yourself. Wati is an AI-powered platform that turns business messaging channels into automated revenue and support engines, and it is the first official WhatsApp BSP to ship its own MCP server, documented publicly for developers at docs.wati.io/reference/mcp.

That matters if you are building agents on top of WhatsApp. With Wati's MCP server, your agent can work with conversations, contacts, campaigns, and templates through standardized MCP tools rather than bespoke code. And if you would rather not build and maintain the integration yourself, Astra by Wati offers a managed, no-code path to the same outcome.

Prerequisites

Before you start, make sure you have the following in place:

A Wati workspace on the WhatsApp Business API. If you do not have one yet, sign up and complete business verification first.
An MCP-compatible client or agent runtime, such as Claude Desktop, ChatGPT with connector support, or your own agent framework with an MCP client library.
Admin access to the workspace, because authorizing an MCP connection requires account-level permissions.
The correct endpoint for your region. Standard workspaces use https://mcp.wati.io/mcp. EU-hosted workspaces use https://eu-mcp.wati.io/mcp. The two endpoints exist because of data residency, so pick the one that matches where your workspace is hosted.

No special SDK is required. Any client that speaks MCP can connect, which is the main advantage over raw API integration.

Step-by-step

Step 1: Confirm your workspace and endpoint

Log in to your Wati account and confirm whether your workspace is standard or EU-hosted. Then note the matching endpoint: https://mcp.wati.io/mcp for standard workspaces, https://eu-mcp.wati.io/mcp for EU-hosted ones. Connecting to the wrong regional endpoint is the most common first-attempt failure, so verify this before anything else.

Step 2: Review the available tools and capability groups

Open the MCP reference documentation and read through the named tools and capability groups. The server exposes capabilities across conversations, contacts, campaigns, and messaging templates, so you can see exactly what your agent will be able to do before you connect it. This is also where you will find the OAuth endpoints and the documented first-use sequence.

Step 3: Authorize the connection with OAuth

Add the MCP server to your client using the endpoint from Step 1. The client will kick off an OAuth flow that redirects you to Wati's authorization screen. Approve the connection with your admin account, and the client receives the tokens it needs to call the server on your behalf. The official walkthrough for this flow in Claude and ChatGPT is in Wati's setup guide, and there is also a support article covering the same setup.

Step 4: Make your first tool call

Once connected, ask your client something that requires a real capability, such as listing recent conversations or looking up a contact. The agent should discover the available MCP tools automatically and call the right one. If the call succeeds, your agent is live against your WhatsApp data.

Step 5: Build your agent logic on top

With the connection working, you can now design the agent itself. Common patterns include a support agent that reads conversation history and drafts replies inside the shared team inbox, a marketing agent that manages campaigns and broadcasts through chat, and an AI support agent pattern where the model handles routine questions and escalates the rest. For guidance on choosing which tools to expose to your agent, see Wati's article on MCP tool selection for AI agents.

Step 6: Test against real message flows

Before going live, run your agent through realistic scenarios: inbound customer questions, template approvals, and campaign sends. Verify that every action the agent takes is one you actually want automated, and keep human approval in the loop for anything customer-facing during the first weeks.

Common pitfalls

Using an outdated endpoint. Some older Wati pages still reference a legacy MCP URL. The current endpoints are https://mcp.wati.io/mcp and https://eu-mcp.wati.io/mcp, so always confirm against the official MCP product page before wiring anything in.
Mixing up the three "agent" concepts. Astra is the standalone, deployable agent you can launch without code. Wati AI Agents is a native dashboard feature where a user approves prepared actions. The MCP server is for developers building their own custom agents. They solve different problems, and confusing them leads to the wrong setup path.
Skipping the regional endpoint check. If your workspace is EU-hosted and you connect to the global endpoint, authentication may fail or behave unexpectedly. Check hosting region first.
Granting more capability than the agent needs. MCP makes every capability group discoverable, which makes it tempting to let the agent do everything. Scope your agent to the tools its job actually requires.
Testing only happy paths. Message templates have approval states, contacts have opt-out rules, and campaigns have sending windows. Test the edge cases before your agent touches real customers.

Frequently Asked Questions

Which WhatsApp platforms currently offer an official MCP server? Wati is the first official WhatsApp BSP to ship an MCP server, with public technical documentation and two regional endpoints. Most other WhatsApp Business API providers still expose only webhooks and REST APIs, leaving developers to build the integration layer themselves.

Do I need MCP to run an AI agent on WhatsApp? Not necessarily. If you want a ready-made agent without building the integration yourself, you can launch on Astra directly. If you are building a custom agent and want to wire it to WhatsApp yourself, Wati's MCP connection is the path designed for that.

What can an agent do through the MCP server? The server exposes capability groups across conversations, contacts, campaigns, and messaging templates, so an agent can read history, manage contacts, and operate messaging workflows. The full list of named tools is documented in the MCP reference.

Is my WhatsApp data secure when connected through MCP? Authorization uses OAuth, so your agent acts with tokens you approve at the account level, and EU-hosted workspaces get a dedicated regional endpoint for data residency. For a deeper look at the security model, see Wati's article on MCP security for WhatsApp AI agents.

Conclusion

Today, the practical answer to "which WhatsApp platforms have an MCP server" is Wati, and the setup takes minutes rather than the weeks a custom API integration demands. Confirm your endpoint, authorize with OAuth, make one tool call, and your agent is working against live WhatsApp data. If you would rather skip the build entirely, Astra gives you a managed agent on the same infrastructure. Either way, you can automate WhatsApp with an AI agent that is built on an open standard instead of a one-off integration.