WhatsApp MCP Server: Connect Claude & ChatGPT (2026)

A WhatsApp MCP server is a tool that connects WhatsApp to AI assistants like Claude, ChatGPT, or Cursor through Anthropic's Model Context Protocol (MCP). Once connected, the AI can read your WhatsApp messages, send replies, manage contacts, look up conversations, and execute complex workflows — all through natural language.

In 2026, WhatsApp MCP servers are evolving fast. The earliest implementations were open-source GitHub projects connecting to WhatsApp Web via Puppeteer or unofficial libraries. The newer wave uses the official WhatsApp Business API and exposes far richer functionality — full lead qualification, campaign management, analytics, and CRM operations.

This guide covers everything: what WhatsApp MCP is, what's available today, the open-source vs commercial options, and how to use it for sales and customer engagement.

What Is a WhatsApp MCP Server?

The Model Context Protocol (MCP) is an open standard introduced by Anthropic in late 2024. It lets AI models (Claude, ChatGPT, Cursor, and others) connect to external tools and data sources through a standardized interface — instead of every AI needing custom integrations for every service.

A WhatsApp MCP server is a server that:

1. Exposes WhatsApp functionality (read messages, send replies, manage contacts) as MCP tools
2. The AI assistant connects to the server using MCP transport (stdio, HTTP, SSE)
3. The AI can then use natural language to perform WhatsApp actions

Once connected, you can ask Claude things like:

• "Show me all my unanswered WhatsApp conversations from this week"
• "Send a follow-up to anyone who said they're interested in coaching but didn't book"
• "Tag all contacts from my Click-to-WhatsApp ad campaign as 'q2-leads' and import them"
• "Pause WhatsApp AI for the next 2 hours, I'm doing manual outreach"

The AI executes these through MCP tool calls. You don't write code — you describe what you want.

Why Use a WhatsApp MCP Server?

Three main use cases:

1. Personal productivity (open-source MCPs)

If you use WhatsApp personally, an open-source MCP server like lharries/whatsapp-mcp lets Claude search your message history, summarize group chats, draft replies in your style, or pull insights from conversations. The AI runs locally — your messages stay on your machine.

2. Sales & lead engagement (commercial MCPs)

If you run a business on WhatsApp Business API, a commercial MCP like SetSmart gives Claude/ChatGPT control over your entire lead pipeline:

• Qualify and book sales calls automatically
• Manage WhatsApp campaigns and broadcasts
• Update CRM and trigger follow-ups
• Pull analytics on conversation rates and bookings

3. Customer support automation

Connect your WhatsApp Business inbox to Claude, then let the AI triage incoming messages, draft responses for human review, and escalate complex issues. You stay in the loop — the AI does the busywork.

Open-Source WhatsApp MCP Servers

As of 2026, there are several open-source WhatsApp MCP servers on GitHub. Most connect to WhatsApp Web (unofficial) via libraries like whatsmeow or Puppeteer. They're great for personal use but not recommended for business use because:

• TOS risk: WhatsApp Web automation violates WhatsApp's Terms of Service. Personal accounts can get banned.
• No official API: No access to WhatsApp Business API features (templates, broadcasts, click-to-WhatsApp ads).
• Scale limits: Designed for one user's personal account, not multi-channel business setups.

Notable open-source options:

lharries/whatsapp-mcp — The most popular WhatsApp MCP server. Built in Go using whatsmeow. Stores messages locally in SQLite. Supports listing chats, searching messages, downloading media, and sending replies. Connects via stdio for Claude Desktop or Cursor.

ericporres/whatsapp-mcp-server — TypeScript-based using Puppeteer. Includes a "chat intelligence processor" that extracts themes and actionable briefings from conversations. More analytical, less operational.

dynamolabs/whatsapp-mcp — Adds group management and richer messaging (reactions, replies to specific messages).

FelixIsaac/whatsapp-mcp-extended — The most feature-rich open-source option. 41 tools including message editing, polls, presence tracking, newsletters, webhooks, and custom nicknames.

jonzarecki/wa-mcp — Lightweight implementation focused on core messaging.

All of these require self-hosting, technical setup (Go or Node.js + WhatsApp Web auth), and they only work with personal WhatsApp accounts via WhatsApp Web.

Commercial WhatsApp MCP: SetSmart

SetSmart is the first commercial WhatsApp MCP server built for businesses. Unlike the open-source options that connect via WhatsApp Web, SetSmart connects through the official WhatsApp Business API — meaning it's safe for business accounts, supports templates and broadcasts, and works at scale.

What's in the SetSmart MCP

SetSmart exposes ~180 MCP tools covering every aspect of a WhatsApp lead engagement workflow:

WhatsApp account management:

• whatsapp.get_status — connection status, provider (Cloud API or 360dialog), AI on/off
• whatsapp.get_profile / whatsapp.update_profile — business profile (about, description, website, photo)
• whatsapp.set_coexistence_auto_disable_ai — pause AI when human takes over
• whatsapp.set_history_import_ai — import past conversations into AI memory

Conversations:

• conversation.list / conversation.search — list + search by status (qualified, booked, etc.)
• conversation.get — full message history of a conversation
• conversation.bulk_tag — tag multiple conversations at once
• conversation.export — pull conversation data to JSON

Contacts:

• contact.find / contact.list — search contacts by name, phone, tags
• contact.import_from_attachment — upload CSV/XLSX and AI maps columns automatically
• contact.add_tag / contact.remove_tag — manage segmentation

Campaigns & broadcasts:

• campaign.list / campaign.create — manage WhatsApp template campaigns
• campaign.send_broadcast — send a WhatsApp template to a contact list
• campaign.get_stats — open rates, click rates, reply rates per campaign

AI control:

• ai.pause_channel / ai.resume_channel — turn the AI on/off for WhatsApp
• ai.update_blacklist — exclude specific contacts from AI handling

Analytics:

• analytics.dashboard — overall stats (qualified, booked, response rate)
• analytics.channel_breakdown — Instagram vs WhatsApp vs Messenger comparison
• analytics.daily_stats — conversation counts and conversion rates by day
• analytics.message_costs — Meta conversation fees breakdown

Bookings:

• calendly.list_event_types / book_calendly_appointment
• ghl.list_calendars / book_ghl_appointment
• iclosed.list_events / book_iclosed_appointment

Other channels (multi-channel agents):

• Instagram tools (instagram.*) — same coverage as WhatsApp
• Messenger tools (messenger.*)

Example: a real conversation with the SetSmart MCP

You're a fitness coach using SetSmart for WhatsApp lead qualification. You open Claude Desktop and type:

"Show me all my qualified leads from the last 7 days who haven't booked a call yet."

Claude calls conversation.search with filters qualified=true, booked=false, channel=whatsapp, date_range=7d. Returns 23 leads. Claude shows you a list with name, phone, last message snippet.

"Send a follow-up to all of them — natural, warm, ask if they want to grab a slot this week."

Claude calls whatsapp.send_message for each contact (or batches via campaign.send_broadcast if a template fits). It uses your historical message style as context. Reports back: "23 follow-ups sent successfully, 0 errors."

"What's my qualification rate this month vs last month?"

Claude calls analytics.daily_stats for both periods. Returns the comparison.

This is a workflow that would take you 2 hours manually. With MCP, it's 3 messages and 30 seconds.

How to Connect SetSmart MCP to Claude

1. Sign up for SetSmart at https://setsmart.io (free 7-day trial)
2. Connect your WhatsApp Business via Meta Cloud API or 360dialog
3. Get your MCP URL in your SetSmart dashboard
4. Add to Claude Desktop:
   • Open Claude Desktop settings
   • Go to MCP servers
   • Add server with your SetSmart MCP URL
5. Authenticate with your SetSmart credentials
6. Use it — Claude now has all 180 tools available

You can also connect to ChatGPT (via custom GPT actions) or Cursor (via MCP config in ~/.cursor/mcp.json).

SetSmart MCP vs Open-Source WhatsApp MCPs

If you're a developer using WhatsApp personally and want to chat with Claude about your messages, open-source MCPs are perfect. If you run a business and want AI to handle real lead engagement, SetSmart's MCP is the only option built for that.

Use Cases for WhatsApp MCP

For coaches and consultants

"Pull all WhatsApp conversations where the prospect mentioned 'budget' but didn't qualify. Show me their objections."

The AI uses conversation.search with text filters, returns the conversations, and summarizes the patterns. You learn what's killing your conversions.

For agencies managing client accounts

"For client X, show me last week's WhatsApp performance. How many qualified leads, booking rate, top objections."

Multi-account agencies can connect each client's SetSmart account and pull reports across all of them in seconds.

For ecommerce stores

"Send a WhatsApp template to everyone who abandoned their Shopify cart in the last 24h. Use the abandoned cart template."

The AI joins Shopify abandoned cart data (via the SetSmart Shopify integration) with the WhatsApp broadcast tool. One sentence, full automation.

For solo founders

"Every morning, summarize last night's WhatsApp conversations and tell me which 3 leads I should personally follow up with today."

Set this up as a scheduled prompt or daily routine. Your AI does the triage, you focus on the highest-value relationships.

What the Data Says

We analyzed 828,761 AI WhatsApp conversations on the SetSmart platform. Key findings relevant to MCP-driven WhatsApp workflows:

• 34% of WhatsApp responders qualify as a real lead (vs 17.77% on Instagram)
• 18.82% of engaged WhatsApp leads book a call — 3x Instagram's rate
• A single follow-up doubles booked calls (+106% among engaged leads)
• WhatsApp leads qualify in a median of 20 messages (~10 back-and-forth exchanges)

These are conversations handled by the SetSmart AI directly — but the same logic applies when you use the MCP layer to orchestrate WhatsApp at a higher level. The AI handles the messaging, the MCP gives you the controls.

For the full benchmark dataset, see our lead response time statistics.

WhatsApp MCP Security Considerations

Whether you use open-source or commercial WhatsApp MCP, watch for:

Message access scope: Some MCPs read your entire message history. Others (like SetSmart) scope access per tool — conversation.list returns a summary, conversation.get returns full text. Use the principle of least privilege.

Token storage: Your WhatsApp Business API token is the equivalent of your password. Open-source MCPs typically store it locally. SetSmart stores it encrypted server-side and exposes only masked previews via tools.

AI memory leakage: If you're using Claude in a shared workspace, conversations passing through the MCP are visible in your chat history. Don't share workspace transcripts with sensitive lead data.

Approval before destructive actions: Good MCPs ask for explicit confirmation before sending broadcasts, deleting contacts, or canceling subscriptions. SetSmart's agent prompt requires confirm=true for destructive operations.

FAQ

What is a WhatsApp MCP server?

A WhatsApp MCP server is a tool that connects WhatsApp to AI assistants like Claude or ChatGPT through Anthropic's Model Context Protocol. Once connected, the AI can read messages, send replies, manage contacts, and execute WhatsApp workflows through natural language. It's the difference between writing code to use the WhatsApp API and just telling the AI what you want.

Is WhatsApp MCP safe?

Open-source WhatsApp MCPs that use WhatsApp Web (unofficial automation) violate WhatsApp's Terms of Service and risk account bans on personal accounts. Commercial options like SetSmart use the official WhatsApp Business API and are safe for business use. Always verify which connection method an MCP uses before relying on it for business.

Can I use WhatsApp MCP with ChatGPT?

Yes. MCP is a protocol standard, not Claude-specific. You can connect WhatsApp MCP servers to Claude Desktop, ChatGPT (via custom GPT actions), Cursor, and any other MCP-compatible client. SetSmart's MCP supports all major MCP clients.

What's the best WhatsApp MCP for business use?

SetSmart is the only commercial WhatsApp MCP server built specifically for business workflows — lead qualification, in-chat booking, campaign management, and analytics. Open-source alternatives (lharries/whatsapp-mcp, FelixIsaac/whatsapp-mcp-extended) are excellent for personal use but not viable for business accounts.

How much does a WhatsApp MCP cost?

Open-source WhatsApp MCPs are free (you self-host on your own machine). Commercial options like SetSmart start at $99/month and include the MCP server, hosting, the WhatsApp Business API integration, and the underlying AI setter that handles lead qualification.

Can a WhatsApp MCP qualify leads automatically?

Open-source MCPs only expose WhatsApp actions — they don't qualify leads on their own. You'd need to write the qualification logic yourself. SetSmart's MCP includes a built-in AI setter that qualifies leads through natural conversation using GPT-level AI, so the MCP layer is for orchestration and analytics, not for the qualification itself.

Does WhatsApp MCP work with Instagram and Messenger too?

The open-source WhatsApp MCPs are WhatsApp-only. SetSmart's MCP includes Instagram MCP tools and Messenger tools alongside WhatsApp, all from one server — useful if you handle leads across multiple Meta channels.