> ## Documentation Index
> Fetch the complete documentation index at: https://developer.chatsyncs.com/llms.txt
> Use this file to discover all available pages before exploring further.

# ChatSyncs MCP — Introduction

> What the ChatSyncs MCP server is and what it lets an AI assistant you already use (Claude, ChatGPT, Claude Code, Codex, Cursor, Gemini CLI, VS Code Copilot) do with your ChatSyncs account: send WhatsApp messages and templates, manage contacts, labels and deals, read conversations. Also covers what you need before connecting, the WhatsApp 24-hour rule that still applies, and what the AI cannot do. Use when a customer asks what MCP is, whether they can run ChatSyncs from ChatGPT or Claude, or what the AI can and cannot do.

The **ChatSyncs MCP server** connects your ChatSyncs account to the AI assistant you already
use: Claude, ChatGPT and others. Once connected, you run your WhatsApp business by asking in
plain English:

> "Send Rahul a message saying his order shipped today."
>
> "Who messaged us in the last hour and hasn't had a reply?"
>
> "Tag everyone who asked about pricing as Hot Lead and assign them to Priya."

The assistant works out which ChatSyncs actions those need, shows you what it's about to do, and
does it on your account.

<Info>
  **MCP** stands for Model Context Protocol, an open standard that lets AI assistants use real
  tools. You don't need to understand it. You add ChatSyncs once in your AI app, paste your
  ChatSyncs API key on a ChatSyncs page, and the assistant gains a set of ChatSyncs abilities.
</Info>

## Connect your AI app

<CardGroup cols={2}>
  <Card title="Claude" icon="message" href="/mcp/apps/claude">Web, desktop and mobile. Easiest; works on the Free plan.</Card>
  <Card title="ChatGPT" icon="comments" href="/mcp/apps/chatgpt">Web. Every action on Business, Enterprise or Edu.</Card>
  <Card title="Claude Code" icon="terminal" href="/mcp/apps/claude-code">One terminal command.</Card>
  <Card title="OpenAI Codex" icon="terminal" href="/mcp/apps/codex">Codex CLI, IDE extension and ChatGPT desktop.</Card>
  <Card title="Cursor" icon="code" href="/mcp/apps/cursor">The Cursor editor's Agent.</Card>
  <Card title="VS Code + Copilot" icon="code" href="/mcp/apps/vs-code">GitHub Copilot Chat.</Card>
  <Card title="Gemini CLI" icon="terminal" href="/mcp/apps/gemini-cli">Google's terminal agent.</Card>
  <Card title="Other apps" icon="puzzle-piece" href="/mcp/apps/other-apps">Windsurf, Zed, Cline, n8n, Make.</Card>
</CardGroup>

Not sure which? [Compare them](/mcp/choose-an-app): if you're not technical, use Claude or
ChatGPT in your browser.

## What you need first

<Steps>
  <Step title="A ChatSyncs account with a connected WhatsApp number">
    The assistant acts on your existing setup. If you haven't connected a number yet, start with
    [Connect your number to ChatSyncs](https://help.chatsyncs.com/docs/connect-whatsapp-number).
  </Step>

  <Step title="Your ChatSyncs API key">
    From the [**API Developer Console**](https://platform.chatsyncs.com/api/developer/console).
    See [Get your API key](/mcp/api-key).
  </Step>

  <Step title="An AI app that supports MCP">
    Any of the apps above. Each has a page with the exact clicks.
  </Step>
</Steps>

## Is this the same as the Developer API?

It's the same power, without the code. The [Developer API](/developer-api/introduction) is for
developers who write a program that calls ChatSyncs. The MCP server is that API already wrapped
up, so an AI assistant can use it for you.

| | Developer API | MCP server |
| - | - | - |
| Who it's for | developers | anyone who can type a message |
| What you build | your own script, n8n flow or app | nothing |
| How you use it | write code, run it | ask the assistant |
| Setup | read the docs, get IDs, write requests | add it in your AI app, paste your key once |

## What the assistant can do

<CardGroup cols={2}>
  <Card title="Messaging" icon="paper-plane">
    Send text, files, images and approved templates. Send messages with tappable reply buttons.
    Check whether a message was delivered or read.
  </Card>

  <Card title="Contacts" icon="address-book">
    Look someone up, add a contact, update their details, read the conversation history, leave
    an internal note.
  </Card>

  <Card title="Organising" icon="tags">
    Create labels and apply them, set custom fields, enrol people in follow-up sequences, update
    deals, resolve or archive conversations.
  </Card>

  <Card title="Team and shop" icon="users">
    Assign a chat to a team member, trigger a chatbot flow, list catalog orders and move them
    to Shipped or Delivered.
  </Card>
</CardGroup>

There are **43 abilities** in total. The full list is on the [Tools reference](/mcp/tools), and
[Things to ask](/mcp/examples) has prompts that work.

## The 24-hour rule still applies

This is WhatsApp's rule, not ChatSyncs', and the AI can't get around it:

* **Within 24 hours** of a customer messaging you, you can reply with anything: free text,
  images, files, buttons.
* **After that**, only an **approved template** will be delivered.

So "message everyone on my list about the sale" needs an approved marketing template. If you
ask for plain text to someone who last messaged three days ago, WhatsApp rejects it. The
assistant knows this rule and normally tells you before trying.

<Warning>
  **Messages sent through the assistant are real.** They reach real customers, use your message
  credits, and can't be unsent. Your AI app asks before anything that sends or changes something.
  Read what it proposes before you approve, and be careful with "always allow" options.
</Warning>

## What it costs

Nothing extra from ChatSyncs. The MCP server uses your existing plan and its limits: your
subscriber cap, your message credits, and Meta's own messaging charges. Your AI app's own plan
is separate (see [which plans work](/mcp/choose-an-app)).

## What the AI can't do

* **Broadcast to a whole list in one step.** There's no bulk-send ability, on purpose. Use
  [Broadcasting](https://help.chatsyncs.com/docs/send-bulk-whatsapp-messages) for
  campaigns.
* **Create or approve message templates.** Make those in ChatSyncs; the assistant can then send
  the approved ones.
* **Change your billing, plan or account settings.**
* **Act on its own.** It only does something when you ask, in a chat where ChatSyncs is turned
  on.

## Frequently asked

<AccordionGroup>
  <Accordion title="Do I need to know how to code?">
    No. With Claude or ChatGPT you click through a few settings and paste your API key into a
    ChatSyncs page. There's no code anywhere in that process.
  </Accordion>

  <Accordion title="Is my API key safe?">
    With browser sign-in, ChatSyncs checks your key when you connect and your AI app stores it.
    The ChatSyncs MCP server keeps no copy. You can cut off every connected app at once by regenerating the key.
    See [Get your API key](/mcp/api-key).
  </Accordion>

  <Accordion title="Can my whole team connect?">
    Yes. Each person connects in their own AI app with an API key from your ChatSyncs account.
    A key carries the full permissions of the account it came from.
  </Accordion>

  <Accordion title="Will the AI message customers without asking me?">
    Not by default. Claude, ChatGPT, Cursor, VS Code and the command-line apps all ask before an
    action that sends or changes something. Deleting a contact and blocking a conversation are
    marked as destructive, so apps warn you more firmly.
  </Accordion>

  <Accordion title="Which AI assistants work with this?">
    Claude, ChatGPT, Claude Code, OpenAI Codex, Cursor, VS Code with GitHub Copilot and Gemini
    CLI have step-by-step guides. Windsurf, Zed, Cline, Make and n8n are covered in
    [Other apps](/mcp/apps/other-apps). Any app that supports remote MCP over HTTP can connect.
  </Accordion>

  <Accordion title="Does this replace my chatbot?">
    No. Your ChatSyncs chatbots keep answering customers automatically, around the clock. The
    MCP server is for *you*: a way to run your account by asking instead of clicking.
  </Accordion>
</AccordionGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.