Skip to main content

​
Overview

Provalytics can be connected to ChatGPT through a custom MCP connector in supported ChatGPT workspaces. Once connected, ChatGPT can work directly with live Provalytics data so teams can ask questions, inspect performance, and prepare answers without exporting files first.

​
What you need

Before starting, confirm you have:
  • a ChatGPT Business, ChatGPT Enterprise, or ChatGPT Edu workspace
  • admin access to that ChatGPT workspace
  • access to the MCP Server Access page in Provalytics
  • a user-specific Provalytics API key
ChatGPT admin labels may vary slightly by workspace configuration or product release, but the overall setup flow is the same.

​
Step 1: Generate your Provalytics API key

In Provalytics:
  1. Open MCP Server Access
  2. Confirm the correct client workspace is selected
  3. Create a key name such as ChatGPT
  4. Click Generate key
Your key is scoped only to that client’s data. MCP Server Access overview

​
Step 2: Enable custom connectors in ChatGPT

In ChatGPT, sign in as a workspace administrator. Then navigate to your workspace admin settings and confirm the features needed for custom connectors are enabled. Depending on your workspace, this may appear under areas such as:
  • Workspace Settings
  • Permissions & Roles
  • Connected Data
  • Apps
Where available, enable:
  • developer mode
  • custom MCP connectors
Some workspaces may already have these enabled.

​
Step 3: Create a new MCP connector

In the ChatGPT workspace admin area:
  1. Go to the section for Apps or Connectors
  2. Choose Create App, Create Connector, or the equivalent option
  3. Select Custom MCP Connector
You will then be prompted for server and authentication details.

​
Step 4: Configure the Provalytics MCP server

Use the following Provalytics server details.

​
MCP server URL

https://mcp.getprova.com/mcp

​
Authentication

Use the user-specific Provalytics API key you generated in Step 1. If ChatGPT asks for an authentication type, use the option that matches token-based authentication, such as:
  • Bearer Token
  • API Token
Important:
  • each user should use their own key
  • keys should be treated as confidential credentials
  • keys are read-only and client-scoped
The Provalytics MCP endpoint is encrypted in transit over HTTPS and supports modern TLS transport.

​
Step 5: Scan available tools

After saving the connector configuration:
  1. Click Scan Tools or the equivalent action
  2. Let ChatGPT connect to the MCP server
  3. Review the discovered tools before publishing
This is the step where ChatGPT detects the Provalytics actions made available through the MCP server.

​
Step 6: Test the connector

Before publishing broadly, open ChatGPT and run a few simple test questions such as:
  • what are my top channels by incrementality this month?
  • what does the latest recommendation suggest?
  • how close did the latest model predictions match actual results?
This confirms:
  • authentication is working
  • the connector can access the correct workspace data
  • the discovered tools are behaving as expected

​
Step 7: Publish the connector

Once validated:
  1. Return to the ChatGPT Apps or Connectors area
  2. Publish the connector
  3. Set visibility for the right users, teams, or workspace groups
Depending on your ChatGPT plan and workspace settings, you may be able to publish it to:
  • specific users
  • teams
  • the full workspace

​
User authentication and permissions

Each user should authenticate with their own Provalytics key. The connector will operate using the permissions associated with that user’s credentials and the selected client workspace. Best practice is:
  • one key per user
  • clear naming such as ChatGPT - Jane Smith
  • revoke keys that are no longer needed
  • test with a small group before wider publishing

​
What ChatGPT can help with

Once connected, ChatGPT can help your team:
  • answer live reporting questions faster
  • summarize findings for internal or client meetings
  • inspect recommendations and forecast outputs
  • compare modeled performance across channels
  • stay inside existing workflows instead of jumping between tools