WhatsApp MCP: Run WhatsApp from Claude (Read, Send, and Schedule in 2026)

You open WhatsApp to "quickly check something" and forty unread threads later you have lost twenty minutes and still missed the one message that actually mattered. The problem was never that you needed to type faster. It was that nothing could read the pile for you and tell you what needs a human.

That is what changed. Blueticks ships an MCP server that connects Claude to your real WhatsApp account, and the first thing it unlocks is not sending, it is reading. Claude can scan your recent chats, summarize a client's entire history, flag what needs your attention, and draft the reply for your approval. Then, when you are ready to act, the same connection sends and schedules from your own number. This guide covers the no-code path (Claude Desktop) and the developer path (Claude Code and the /v1 API).

What is the WhatsApp MCP and why does it matter now?

The WhatsApp MCP is a Model Context Protocol server from Blueticks that lets an AI assistant like Claude read and operate your real WhatsApp account. MCP is the open standard, introduced by Anthropic in late 2024, that lets AI models call external tools through a shared interface. The Blueticks MCP exposes WhatsApp read, send, schedule, and campaign actions as those tools.

Before MCP, connecting an AI to WhatsApp meant writing a custom integration: API client, auth handling, error retries, the lot. MCP collapses that into a config block. Once the Blueticks server is registered, Claude sees a set of WhatsApp tools and decides which to call based on what you ask. You type "what did Noam say about the demo?" and Claude reads the thread and answers, no app-switching required.

It matters now because attention, not typing, is the bottleneck. The MIT Lead Response Management study found you are 21 times more likely to qualify a lead when you reply within five minutes instead of thirty, and the threads that decide deals are buried under group chatter and "thanks!" replies. An assistant that reads the pile and surfaces the one message that needs you is the difference between catching that window and missing it. The @blueticks/mcp package (currently v0.3.0) is published on npm and runs with a single npx command, so there is nothing to clone or compile.

How does Claude work as your WhatsApp chief of staff?

Claude becomes your WhatsApp chief of staff by reading your chats for you. With the Blueticks MCP connected, it can list and search your recent conversations, read the messages inside any thread, and look up contacts, then turn all of that into a digest, a triage list, or a drafted reply you approve before anything sends.

Four read-path requests carry most of the daily value:

• "Summarize my day on WhatsApp." Claude pulls your recent chats with whatsapp_chats, reads the new messages, and hands you a short digest: who messaged, what they wanted, what is still open. You get the gist of forty threads in one paragraph.
• "What needs my attention right now?" Claude scans recent conversations and triages them, separating the threads waiting on a reply from the noise. Instead of scrolling, you get a ranked shortlist of what actually needs a human.
• "Draft a reply to my client at Hadar Tours." Claude reads the full thread for context, then writes a reply in your voice and shows it to you. Nothing leaves your account until you say go. You go from "I need to get back to them" to a polished draft in seconds.
• "Summarize this client's entire interaction history." Claude reads the whole conversation, not just the latest message, and gives you the arc: what they bought, what they complained about, what they are waiting on. It is a one-line CRM lookup without a CRM.

These all run on read-only tools. whatsapp_chats reads your chats and the messages inside them, with list, search, and per-thread message history. whatsapp_contacts reads the contacts your engine knows. Claude never sends a thing in any of the above; it reads, it summarizes, it drafts. The send is a separate, deliberate step you trigger.

In practice it reads like a Monday-morning conversation. You type "catch me up on WhatsApp since Friday and tell me what needs a reply," and Claude comes back with the shortlist: a client wants to move a call, a lead asked about pricing, a supplier confirmed a delivery, the rest is handled. You say "draft a reply to the client confirming Wednesday at 2pm," Claude reads the thread for tone, drafts it, and waits for your nod. As one early operator put it, "I used to spend the first hour of my day just reading WhatsApp. Now Claude reads it and I spend ten minutes deciding." That read-triage-draft loop is the daily driver, and it is why the MCP earns a place in your morning routine rather than just your campaign stack.

What can you actually do with it (the full tool set)?

The Blueticks MCP gives Claude a focused set of WhatsApp tools, not a magic "do anything" button. It covers reading, sending, scheduling, contacts, groups, audiences, campaigns, webhooks, and engine status. Every action runs through your own connected WhatsApp session, so recipients see a normal message from you.

Here is the honest inventory, taken straight from the package:

The split is clean: the top two tools are your read path (the chief-of-staff layer), and the rest are your write path. Sending and scheduling are first-class, the same engine that powers Blueticks scheduling and campaigns does the delivery. For how that scheduling layer works on its own, see our guide to scheduling WhatsApp messages. What it does not do: give you a second number, bypass WhatsApp's own rules, or message people who never opted in for you.

How do you act on it: send and schedule from a sentence

Once Claude has read the situation, acting on it is one more sentence. You name the people and the time; Claude anchors the time to your timezone, drafts the message for approval, then sends now or schedules for later. Scheduling is a core capability, not an afterthought.

Say you finish your morning triage and type:

"Schedule a WhatsApp to my 20 VIP customers tomorrow at 9am letting them know early access opens Monday."

Behind the scenes, Claude does roughly this:

1. Calls whatsapp_utils to get your current date, time, and timezone, so "tomorrow at 9am" resolves to a correct timestamp instead of the server's clock.
2. Pulls the VIP list, either from a saved audience via whatsapp_audiences or from contacts you name.
3. Drafts the message, shows it to you, and waits for your go-ahead.
4. Schedules each send with whatsapp_messages and a send_at time, or runs a paced campaign through whatsapp_campaigns for per-contact variables like {firstName}.

Later you can ask "show me my scheduled WhatsApp messages" and Claude lists what is pending via whatsapp_scheduled_messages. Change your mind? "Reschedule the VIP message to 11am" or "cancel it" both work, because that tool supports update and delete. The read path tells you what to do; the write path does it on your schedule.

How do you connect the WhatsApp MCP to Claude in about 2 minutes?

To connect the Blueticks MCP to Claude, create a Blueticks API key, then add a small server entry to your Claude config that runs npx -y @blueticks/mcp with that key in the environment, then restart Claude. The whole thing takes about two minutes and needs no cloning or building.

Step one is the same for everyone: get a key. Sign in to the Blueticks API dashboard, choose Create key, and copy the value immediately. Keys look like bt_live_... and are shown only once, so paste it somewhere safe. Treat it like a password; it can read and send from your account.

Claude Desktop (no code). Open your Claude Desktop config file. On macOS that is ~/Library/Application Support/Claude/claude_desktop_config.json. Add this block, paste your key, and restart Claude Desktop:

{
  "mcpServers": {
    "blueticks": {
      "command": "npx",
      "args": ["-y", "@blueticks/mcp"],
      "env": {
        "BLUETICKS_API_KEY": "bt_live_your_key_here"
      }
    }
  }
}

Claude Code (developers). The same server entry goes in your .mcp.json (project) or ~/.claude.json (global):

{
  "mcpServers": {
    "blueticks": {
      "command": "npx",
      "args": ["-y", "@blueticks/mcp"],
      "env": {
        "BLUETICKS_API_KEY": "bt_live_your_key_here"
      }
    }
  }
}

Restart Claude Code after saving. (Claude Code also supports adding servers from the command line if you prefer that workflow; the JSON block above is the method Blueticks documents and the one that always works.) The first time a WhatsApp tool runs, npx pulls the package automatically, so the only prerequisite is Node.js 20 or newer.

One thing to confirm before you start: your WhatsApp engine has to be connected to Blueticks first, the same connection that powers scheduling and campaigns in the app. Once it is linked, ask Claude "is my WhatsApp connected?" and it will call whatsapp_engine and tell you.

For developers: the same power through the /v1 API

Developers get the identical capabilities without Claude in the loop by calling the Blueticks REST API directly. The MCP server is a wrapper over the same /v1 endpoints, so anything Claude can do, your own code can do, with full control over retries, idempotency, and webhooks. The base is api.blueticks.co with documented endpoints under /v1/.

The auth model is the same key you used for the MCP. Pass it as a bearer token:

Authorization: Bearer bt_live_your_key_here

From there you have POST /v1/scheduled-messages to send or schedule, GET /v1/scheduled-messages/:id to check status, POST /v1/campaigns to start a paced campaign, and POST /v1/webhooks to register signed event callbacks. The MCP and the API are two front doors to one engine. Use the MCP when a human is driving in natural language, and the API when a backend service needs deterministic, programmatic sends. A practical pattern is to prototype the workflow conversationally in Claude, then port the proven flow into a scheduled job against /v1. Full reference, SDKs for five languages, and the cURL quickstart live at dev.blueticks.co.

MCP vs the WhatsApp Business API: when to use which?

Use the WhatsApp MCP when you want Claude to read and operate your own existing WhatsApp number with a personal feel and minimal setup. Use the official WhatsApp Business Platform (the Cloud API) when you need Meta-verified, template-based messaging at very high volume with formal opt-in management. They solve different problems.

Neither one lets you spam. The Cloud API enforces templates and opt-in at the platform level; the MCP runs through your personal account, so WhatsApp's normal anti-abuse limits still apply to you. The MCP wins on the read-and-reply workflow and speed-to-value. The Cloud API wins when you need Meta's enterprise scale and are prepared for the verification overhead. Many teams start on the MCP and move to the Cloud API only when outbound volume forces it. For a deeper look at running structured sends, see our WhatsApp campaign management guide.

What are the limits and good practices?

The core rule is simple: the MCP reads and automates your real WhatsApp account, so every WhatsApp rule that applies to you still applies through the MCP. Read freely, but message only people who expect to hear from you. Three practices keep you safe:

• Read first, send deliberately. The read tools are safe to use constantly, that is the point. Keep sending as a separate, approved step, and have Claude show you the draft and recipient count before anything dispatches.
• Own number, real consent. The MCP sends from your account to contacts who opted in. It is outreach automation, not a cold-blast tool. A list of strangers will get your number flagged regardless of how you send.
• Pace your sends. For bulk, use whatsapp_campaigns rather than firing hundreds of messages in seconds. Start with a small batch, watch the delivery acknowledgements, then scale up.

One practical note: a freshly queued message has no WhatsApp wire key until it actually dispatches, so delivery and read receipts populate a moment after sending, not instantly. Ask Claude to check status rather than assuming a send confirmed the instant you typed it.

Frequently asked questions

Can Claude read my WhatsApp chats, or only send messages? It can read. With the Blueticks MCP connected, Claude can list and search your chats, read the messages inside any thread, and look up contacts, which is what powers summaries, triage, and drafted replies. Sending and scheduling are separate actions you trigger.

Do I need to code to use the WhatsApp MCP? No. The Claude Desktop path is a single JSON config block and an API key, with no programming required. Developers who want programmatic control can use Claude Code or call the /v1 REST API directly, but it is optional.

Does the MCP work with my own WhatsApp number? Yes. Reading and sending both go through your own connected WhatsApp session via the Blueticks engine, so recipients see a normal message from you, not from a third party. You need your WhatsApp engine linked to Blueticks first.

Can Claude schedule WhatsApp messages for later, not just send now? Yes. Scheduling is a core capability. You can schedule a send for a future time, and list, reschedule, or cancel queued messages in plain English. Claude checks your current time and timezone first so relative times like "tomorrow at 9am" land correctly.

Will using AI on WhatsApp get my number banned? Not if you follow WhatsApp's rules. The MCP does not bypass anything; it reads and sends as you. Reading is low-risk. For sending, message opted-in contacts, pace bulk sends through campaigns, and avoid blasting strangers, and you are operating exactly as a careful human would.

Connect Claude to your WhatsApp today

Stop starting your day by scrolling. Create a Blueticks API key, drop the config block into Claude, and ask it to read your WhatsApp and tell you what needs you. Get your API key and the full MCP setup at dev.blueticks.co, then have Claude run your inbox.