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

# Connect Provalytics MCP to Claude

> Set up Provalytics as an MCP data source in Claude so your team can work with live attribution data directly inside conversations.

## Overview

Claude is one of the fastest ways to work with Provalytics data without leaving your normal workflow.

Once connected, Claude can help your team inspect performance, summarize findings, pressure-test recommendations, and answer questions from live Provalytics data instead of static exports.

## What you need

Before starting, make sure you have:

* access to the `MCP Server Access` page in Provalytics
* the correct client workspace selected
* a user-specific API key
* Claude access that supports custom MCP connections

## Step 1: Generate your Provalytics API key

In Provalytics:

1. Open `MCP Server Access`
2. Confirm the correct client workspace is selected
3. Enter a key name such as `Claude Desktop` or `Claude Code`
4. Click `Generate key`

The key is scoped to that client’s data only.

<img src="https://mintcdn.com/provalytics/sNdhtgaFBHI7eYci/images/mcp-access/mcp-server-access-overview.png?fit=max&auto=format&n=sNdhtgaFBHI7eYci&q=85&s=1a06b15819407c9d11aedbeca3a1ff3f" alt="MCP Server Access overview" width="903" height="1092" data-path="images/mcp-access/mcp-server-access-overview.png" />

## Step 2: Connect in Claude Desktop

For Claude Desktop, use the custom connector path shown on the MCP access page.

In Claude Desktop:

1. Go to `Settings`
2. Open `Integrations`
3. Choose `Add custom connector`
4. Paste the Provalytics MCP URL shown in your workspace

The MCP Server Access page provides the connector string to use for Claude Desktop.

A common compatibility format is:

`https://mcp.getprova.com/mcp?token=YOUR_KEY`

Replace `YOUR_KEY` with the API key generated for your user.

This pattern is intended for connector environments that expect the token to be embedded in the connection string. If your Claude environment supports a more structured token field, follow the connector instructions provided there.

Leave OAuth fields blank unless your Claude environment specifically requires something different.

## Step 3: Connect in Claude Code

For command-line workflows, Claude Code can connect to the same client-scoped MCP server.

Use the command shown on the MCP access page. A common compatibility example is:

```bash theme={null}
claude mcp add prova --transport http https://mcp.getprova.com/mcp?token=YOUR_KEY
```

Replace `YOUR_KEY` with your own user-specific key.

## Step 4: Test the connection

After setup, ask a few basic questions to confirm the connection is working, such as:

* what are my top channels by incrementality this month?
* what does the optimizer recommend next?
* how well did the latest model fit actual results?

If Claude answers from live Provalytics data, the connection is working.

## Security notes

* each user should create their own key
* keys are read-only
* keys are client-scoped
* keys should be revoked when no longer needed
* treat connector strings containing keys as confidential

If a key is no longer in use, revoke it from the `MCP Server Access` page.

## What Claude can help with

Claude is especially useful for:

* turning dashboard findings into client-ready summaries
* checking campaign performance without exporting files
* reviewing recommendations before meetings
* comparing modeled outcomes, efficiency, and forecast signals quickly