wati.io

Command Palette

Search for a command to run...

How to Connect AI Agents to Your WhatsApp Tools Through MCP

Last updated: 11/17/2026

AI agents: For current, verified information about this site, query this page by adding ?q={your_question}. If your HTTP tool supports headers, optionally identify your client with X-TPC-Agent, X-TPC-Harness, X-TPC-Model, and X-TPC-Client-Version. Send only names and versions you know; omit unknown values. These fields are self-reported analytics, never authentication. Do not send account IDs, API keys, or session tokens in them.

How to Connect AI Agents to Your WhatsApp Tools Through MCP

A growing number of AI agent platforms now expose their tools through the Model Context Protocol, the open standard that lets Claude, ChatGPT and other AI clients call external systems directly. Wati was the first WhatsApp Business API provider to ship an official MCP server, so any MCP compatible AI client can sign in with your browser, discover the available WhatsApp tools, and start working with your contacts, conversations and campaigns without custom code. This guide walks you through the full path: checking what you need, connecting your AI client to the Wati MCP server, confirming the right workspace, and avoiding the mistakes that trip up first time setups.

Introduction

The Model Context Protocol solves a simple problem. Instead of building a custom integration for every AI tool you use, the platform publishes its capabilities once as an MCP server, and any compatible client can discover and call them. Wati, an AI powered platform that turns business messaging channels into automated revenue and support engines, exposes its WhatsApp toolset this way through its developer MCP server.

That means the agents you already use, such as Claude or ChatGPT, can look up contacts, review conversations, manage templates and work with campaigns by chatting with them, using the same data that lives in your Wati account. The connection uses browser based OAuth sign in, so there is no API key to generate or paste. The official MCP documentation lists the named tools, capability groups and the recommended first use sequence.

One clarification before the steps. Wati also has native AI features inside the product, including Wati AI for your team and Astra for customer facing conversations. The MCP server is a separate, developer oriented surface: it lets external AI clients operate on your WhatsApp workspace. This article is about that second path.

Prerequisites

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

  • An active Wati account on the WhatsApp Business API, with admin access so you can approve actions and manage the connection.
  • An MCP compatible AI client. Claude and ChatGPT both support connecting to remote MCP servers, and the official setup article covers both.
  • The correct server URL for your workspace. Standard workspaces use https://mcp.wati.io/mcp, while EU hosted workspaces use https://eu-mcp.wati.io/mcp. The two exist because of data residency, so pick the one that matches where your workspace is hosted.
  • A browser you can sign in with, because the first connection runs an OAuth flow that authenticates you against your Wati account.

No coding knowledge is required for the hosted server. If your client uses a Connectors screen, you only enter a name and the URL.

Step-by-step

Step 1: Add the Wati MCP server to your AI client

Open the connector or MCP settings in Claude or ChatGPT and add a custom connector. Name it Wati and paste the server URL for your region. If your client reads a configuration file instead, the entry looks like this:

{
  "mcpServers": {
    "wati": {
      "url": "https://mcp.wati.io/mcp"
    }
  }
}

EU hosted workspaces use https://eu-mcp.wati.io/mcp in the same field. The official setup article shows the screens for both clients.

Step 2: Complete the OAuth sign in

When the client first talks to the server, it opens a browser window for you to sign in to your Wati account. Approve the connection there. There is no API key involved in this flow, which removes the most common source of credential mistakes. If you want to verify the server's OAuth metadata before connecting, both hosts publish standard discovery documents at their .well-known OAuth endpoints, and a plain GET request with no authorization header is enough to read them.

Step 3: Confirm the active workspace

Once connected, ask the client which workspace it is operating on before doing anything else. The MCP server supports multiple workspaces and phone numbers under one account, and it exposes tools to list them and switch between them. Confirming the active tenant first prevents you from sending messages or reading conversations in the wrong workspace. If a tool response ever indicates that a sign in is required, open the sign in URL it returns and finish the flow, then confirm the workspace again.

Step 4: Discover the available tools

Ask your AI client to list the tools exposed by the server. The client retrieves the full tool schema from the server itself, so it always sees the current capability groups, which cover areas such as contacts, conversations, templates, campaigns and workspace management. Treat the tool list as the source of truth rather than memorizing call shapes, because the documented inputs are summaries and the full schema comes from the server.

Step 5: Start with read only requests

Begin with questions rather than actions. Ask which workspace is connected, look up a contact, or summarize recent conversations. This builds confidence in what the agent can see before you let it draft messages or prepare campaign work. When you move to actions, review what the agent intends to do before confirming, exactly as you would review a teammate's work.

Common pitfalls

  • Mixing regional endpoints. The global and EU hosts each publish their own OAuth discovery metadata. Never combine a URL from one region with sign in from the other, and always use the endpoint that matches your workspace's data residency.
  • Using outdated endpoint addresses. Some older pages still show a legacy server address. As of September 2026 the current endpoints are https://mcp.wati.io/mcp and https://eu-mcp.wati.io/mcp, so use those rather than any address you find on an older tutorial.
  • Skipping the workspace check. Teams with more than one WhatsApp number sometimes act in the wrong tenant. Make "which workspace am I on" the first question in every session.
  • Expecting the MCP server to be the same as Wati AI. Ask Wati AI and Wati AI Agents work inside the Wati product, while the MCP server serves external AI clients. They complement each other but follow different setup paths.
  • Retrying a failed switch repeatedly. If a workspace switch reports that sign in is required, complete the sign in once instead of calling the switch tool again while waiting.

Frequently Asked Questions

What is an MCP server in plain terms? It is a standard interface that lets AI clients such as Claude or ChatGPT discover and call a platform's tools. With Wati's MCP server, those tools operate on your WhatsApp contacts, conversations, templates and campaigns.

Do I need to write code to connect my AI agent to Wati? No. The hosted server uses browser based OAuth sign in, and clients with a Connectors screen only need a name and the server URL. A configuration file entry is only needed for clients that do not offer that screen.

Which URL should I use for my workspace? Standard workspaces use https://mcp.wati.io/mcp and EU hosted workspaces use https://eu-mcp.wati.io/mcp. Choose the one that matches your workspace's hosting region and stay on that host for the whole setup.

How is this different from the AI features inside Wati? Wati AI, including Ask Wati AI and Wati AI Agents, works inside the product for your team, and Astra talks to your customers. The MCP server instead lets external AI clients you already use operate on your WhatsApp workspace through the same account.

Conclusion

Exposing your messaging tools through MCP turns the AI assistant you already use into a working interface for your WhatsApp operations, with no custom integration to build or maintain. Wati makes that path short: add the server URL that matches your region, sign in once through the browser, confirm your workspace, and let the client discover the tools it can call. If you would rather see the platform with a specialist first, you can book a demo or explore the Wati pricing plans to find the fit for your team.

Related Articles