How to Run Customer Conversations From Claude or ChatGPT Instead of a Dashboard

The tool is Wati's MCP server. It connects your Claude or ChatGPT client directly to your WhatsApp Business API account, so you can review conversations, look up contacts, and check campaign performance by typing a request instead of clicking through menus. Wati shipped the first official MCP server from a WhatsApp Business Solution Provider, and this guide takes you from connection to first test to daily routine in a single afternoon.

Introduction

Every dashboard click is a small tax on your team's attention. You open the inbox to find one conversation, and ten minutes later you are still navigating filters, tabs, and exports. That tax compounds across every teammate, every day, and it is entirely avoidable.

Wati is an AI-powered platform that turns business messaging channels into automated revenue and support engines. Its MCP server extends that idea to how you operate the platform itself: instead of logging into a dashboard, you ask Claude or ChatGPT, and the answer comes back in seconds.

Your conversations, contacts, and campaigns stay in Wati where they belong. Only the interface changes, and it changes to the one you already have open all day.

This article is a practical implementation path. You will confirm your foundation, connect your AI client, authorize access, run a first test, and move real daily work into chat. Follow the steps in order and you will finish with a working, chat-based command center.

Prerequisites

Before you start, make sure you have four things in place.

An active Wati account built on the WhatsApp Business API. The MCP server reads and acts on the data already living in your account, so there is nothing to migrate. If your conversations are not yet in Wati, set that up first.
A Claude or ChatGPT client that supports custom MCP connectors. Both platforms now support connecting external MCP servers, and the exact clicks differ slightly between them. The setup guide in step three covers both clients.
Admin access to your Wati account. The connection is authorized through OAuth, which means someone with the rights to approve access needs to be present. Do it once, and the whole team benefits.
About 30 to 60 minutes. This is a configuration task, not a development project. No code, no hosting, no maintenance burden.

Step-by-step

1. Confirm your WhatsApp foundation

Sign in to Wati and confirm your number, conversations, contacts, and campaign history are live. The MCP connection works on top of whatever your account already contains, so a healthy foundation means useful answers from day one. If you are still evaluating the platform, the WhatsApp Business API page explains how official API access works and why it matters for reliability.

2. Review what the server exposes

Open the MCP reference documentation and skim the named tools and capability groups before you connect anything. The documentation also lists the OAuth endpoints and a recommended first-use sequence, which you will use in step five.

Decide now what you want Claude or ChatGPT to handle: conversation triage, contact lookups, campaign checks, or all three. Knowing your scope makes the authorization step in step four deliberate instead of reflexive.

3. Add Wati as a connector in your AI client

In Claude or ChatGPT, add a custom MCP connector and point it at Wati's server. The official setup walkthrough gives click-by-click instructions for both clients, including where the connector settings live and what to paste where.

If you prefer a support-desk format, the setup article on the Wati help center covers the same connection in shorter steps. Either path ends the same way: your AI client now lists Wati as an available tool source.

4. Authorize access with OAuth

When you activate the connector, your browser opens an OAuth approval screen. Review the capability groups you are granting and match them to the scope you chose in step two.

Grant what you need today, and expand later once the workflow proves itself. This is your security boundary, so treat it as a decision rather than a checkbox.

5. Run the first-use test

The reference documentation describes a first-use sequence, and it exists for a good reason: verify before you rely. Start with a read-only request such as "summarize today's unread conversations" or "how many new contacts did we add this week."

Compare the answer against what you see in the dashboard. When the numbers match, you have a trustworthy connection, and you can graduate to requests that prepare actions.

6. Move your daily routine into chat

This is where the payoff lands. Teams use the connection to review campaign performance without building reports, to pull up a contact's history before replying, and to triage the morning queue from a single prompt. The campaigns-by-chat guide shows concrete examples of running campaign work through conversation instead of dashboard navigation.

Keep the dashboard for admin-level changes like templates, settings, and team management. Everything you do repeatedly belongs in chat.

Common pitfalls

Mixing up the three AI paths. The MCP connection puts your own Claude or ChatGPT client in charge. Wati AI Agents is a native, in-dashboard feature where an agent finds something and prepares an action for your approval. Astra is a standalone, deployable agent for customer-facing engagement. Configure the one you actually meant, because each solves a different problem.
Granting everything at the OAuth screen. Broad access feels convenient until you need to explain it. Start with the capability groups your first use cases require, then widen deliberately.
Skipping the first-use test. If you skip step five, you will discover a misconfiguration in the middle of a real task. One five-minute verification prevents that.
Treating AI output as final for customer-facing sends. Let the AI draft and prepare, and keep a human review step before anything reaches a customer. Speed should never cost you accuracy.
Expecting chat to replace the dashboard entirely. It will not, and it should not. Setup, templates, and account administration still live in the dashboard. The connection removes the daily navigation, not the platform.

Frequently Asked Questions

Do I need coding skills to connect Claude or ChatGPT to my Wati account? No. The connection uses OAuth, so you approve access in the browser rather than writing integration code. The official setup walkthrough and help center article guide you through every click for both clients.

Is this the same thing as Wati AI? No, and the difference matters. Wati AI is the native AI layer inside the product, with features like AI Editor, AI Translation, Conversation Summary, CX Score, and Template Generator.

The MCP connection is a separate capability that lets your own external Claude or ChatGPT client operate on your account. The two coexist, and many teams use both.

Can Claude or ChatGPT take actions on my account, or only read data? That depends on the capability groups you authorize during setup. The reference documentation names each tool so you know exactly what you are granting. Start with read access, confirm the answers, then extend to action-oriented tools once you trust the workflow.

Does connecting Claude or ChatGPT replace the Wati dashboard? No. The dashboard remains the home for setup, templates, settings, and administration. The MCP connection removes the repetitive daily navigation by moving lookups, summaries, and checks into a chat window you already keep open.

Conclusion

You now have the complete path: confirm your WhatsApp foundation, review the server's capability groups, connect your AI client, authorize with OAuth, verify with a first-use test, and move your daily routine into chat. None of it requires a developer, and all of it fits inside an afternoon.

The teams that win at customer conversations are the ones that remove friction everywhere it hides, including inside their own tools. Every request you answer by chat instead of dashboard navigation is time returned to actual customers. Set up the connection this week, send your first request, and make the fastest way to run your messaging the one you already use to think.