> For the complete documentation index, see [llms.txt](https://docs.perisclaw.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.perisclaw.com/use-with-claude-and-other-ai/mcp-server.md).

# Perisclaw as an MCP server

You can connect your Perisclaw WhatsApp to an external AI client — **Claude Desktop, Claude.ai, Cursor, Codex**, and others — and then read and act on your WhatsApp from that client. Ask Claude to *"summarize my unread chats,"* *"reply to Sarah,"* or *"add everyone in the Vendors group to a broadcast,"* and it works through your assistant.

This is the mirror image of [connecting apps](/skills-and-integrations/connecting-apps.md): there, Perisclaw uses your other tools; here, your other tools use Perisclaw. It works because Perisclaw can act as an **MCP server** — a standard way to expose your WhatsApp as a set of tools any MCP-compatible AI can use.

{% hint style="info" %}
Everything happens on **your own** WhatsApp, through your private server. No second copy of your account is created, and the same sending safeguards that apply in chat apply here too.
{% endhint %}

## Connect a client

The smoothest experience is in clients that support one-click connectors (like Claude Desktop and Claude.ai):

{% stepper %}
{% step %}

### Add the connector

In your AI client, add a new connector / MCP server and paste your Perisclaw MCP address (from the **MCP** page in your dashboard).
{% endstep %}

{% step %}

### Sign in

The client opens a Perisclaw login in your browser. Because you're already signed in to the dashboard, there's nothing new to set up — you just see a consent screen.
{% endstep %}

{% step %}

### Choose what it can do

Pick an access level — **Read only** or **Read + write** — then approve. That's it; the client can now use your WhatsApp.
{% endstep %}
{% endstepper %}

## Choose the right access level

| Access level     | What the client can do                                                                                      |
| ---------------- | ----------------------------------------------------------------------------------------------------------- |
| **Read only**    | Search and read your chats, contacts, and groups. Can't send or change anything. The safe default.          |
| **Read + write** | Everything in Read only, **plus** sending messages, reacting, managing chats, groups, labels, and contacts. |

{% hint style="warning" %}
Start with **Read only** unless you specifically want the client to send messages or make changes. You can always reconnect with a higher level later, and you can revoke access at any time from the dashboard.
{% endhint %}

## What you can do from your AI client

* **Catch up** — *"Summarize everything I missed in my work groups today."*
* **Search** — *"Find the address Priya sent me last month."*
* **Reply** (with Read + write) — *"Reply to Sarah that Tuesday at 3 works."*
* **Organize** (with Read + write) — *"Label all the unanswered client chats 'follow up.'"*

## Staying in control

* **Scoped to you** — a connection only ever reaches your own WhatsApp; it can't touch anyone else's.
* **Sends are still gated** — messages sent from an AI client go through the same per-chat permission and [mode](/security-and-control/chat-modes.md) checks as everything else.
* **Revoke anytime** — remove a connection from the dashboard and the client immediately loses access.

For the bigger picture on what your assistant can and can't touch, see [Privacy & security](/security-and-control/privacy-and-security.md) and [Permissions & access](/security-and-control/permissions-and-access.md).


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.perisclaw.com/use-with-claude-and-other-ai/mcp-server.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.