Using the MCP Page in Your PMA Hub

Using the MCP Page in Your PMA Hub

Info
The MCP page in your PMA Hub is where you connect AI assistants to your marketing data, manage those connections, and see how much of your plan's request and row allowance each connection uses. This article walks through everything the page does. For a full overview of the MCP server (what it is, the available tools, and worked examples), see the Power My Analytics MCP Server Guide.
Notes
The PMA MCP server is in Beta and available to all PMA users. There is no early-access request and no allow-listing; you connect directly from the MCP page. MCP usage is included with your PMA plan and draws from the same shared per-plan limit pool as the Power My Analytics API (each plan includes an API rate limit, a request quota, and a row quota per billing cycle; on the Custom plan the request and row quotas are negotiated with your sales representative). See Get Started with Power My Analytics API for the current per-plan limits.

What the MCP Page Is

The MCP page is the home base for Power My Analytics' Model Context Protocol (MCP) server inside your Hub. From one screen you can:
  • Connect an AI assistant (such as Claude Desktop, Claude Code, or ChatGPT) to your Hub.
  • See every application currently connected to your organization.
  • Revoke any connection you no longer want.
  • See how many requests and rows each connection has used this billing cycle.
Once an assistant is connected, you can ask questions about your warehoused marketing data in plain language instead of building a report first.

Finding the MCP Page

Open your Hub at hub.powermyanalytics.com. In the left navigation, open the MCP & API group and click MCP (it carries a Beta badge). The page lives at /organizations/<your-org-id>/mcp. The REST API page, where you manage your organization's API keys, sits in the same group.
Any Hub user with access to the organization can open the MCP page and connect an AI assistant. You do not need an admin role.

The MCP page in the PMA Hub

Connecting an AI Assistant

  1. On the MCP page, click + Connect AI.
  2. Choose your assistant from the menu: Claude Desktop, Claude Code, or ChatGPT. Each option opens a short, step-by-step guide for that client.
  3. Follow the guide to add the PMA connector and complete the sign-in (OAuth) flow using your existing PMA login (Google single sign-on or email and password).
When you connect through the sign-in flow, your access token is created automatically. There is no manual API token to generate or paste, and you do not need to contact support to be provisioned.
If you first connected before PMA introduced per-connection tokens and per-user usage visibility, sign out of the connector and reconnect once so those changes take effect.
Notes
Using a different MCP-compatible client (such as Cursor) or building your own integration? Browser-based clients connect with the same server URL and OAuth sign-in described in the Power My Analytics MCP Server Guide. For a backend service with no browser, use the Power My Analytics REST API with an organization API key from MCP & API > REST API instead; API keys can't be used to connect to the MCP server. See Get Started with Power My Analytics API.
The connect guides for each client:

Understanding the Connected Application Table

The Connected application table lists the applications your organization has connected to its MCP server. If a member of the PMA support team connects to your organization to help troubleshoot an issue, that connection is not shown in your table, and its activity is never counted toward your usage or billed to you. The table has these columns:
Column
What it shows
Connected application
The connection's name, such as the AI client (for example, Claude). Older token-based connections show as API Token.
Connected by
The person who made the connection, shown by name or email. Hover over a name to see the person's email. Manual appears when no person is recorded for the connection.
Requests
How many requests this connection has made in the current billing cycle.
Rows
How many rows of data have been returned to this connection in the current billing cycle.
Date created
When the connection was first made.
Last active
The most recent time the connection made a request.
Trash icon
Revokes the connection (see below).

Use the Search box and the Connected by filter (Anyone, or one person) to narrow a long list, and click any column header to sort by it. Once a connection has been revoked, Active and Revoked tabs appear above the table so you can switch between current and revoked connections.
This table is also how you confirm a connection actually registered. If you completed sign-in but your AI client says it cannot connect, check here: a row for your client means the connection registered on PMA's side (the issue is on the client or the network), and no row after several tries means the sign-in did not complete (reconnect, or contact support).

Hover-Over Details

Several parts of the MCP page show extra information when you hover over them:
  • Requests or Rows column header: the billing cycle the numbers cover (for example, "Usage for billing cycle Sep 24 – Oct 23, 2026").
  • A name in Connected by: that person's email address (shown when the name isn't already the email).
  • The trash icon: Revoke connection.
On the REST API page, the Requests and Rows headers show the same billing-cycle tooltip, each key shows Click to copy, and the trash icon shows Delete API key. On your Hub's Overview page, hover the info icon next to each summary card (Sources, Accounts, Data Sync Calls, Members) to see what it counts.

Revoking a Connection

  1. In the Connected application table, find the connection you want to remove.
  2. At the end of its row, click the trash icon (Revoke connection).
  3. In the confirmation dialog ("Are you sure you want to revoke this connection?"), click Confirm.
Revoking a connection immediately ends that application's access to your Hub through the MCP server. Your Hub and your marketing data are not affected; only that application loses query access. The user can reconnect later from their AI client if needed.
The revoked connection moves to the Revoked tab, and its usage drops out of the table and the MCP requests and MCP rows figures. Requests and rows it already used this billing cycle still count toward your Total requests (MCP & API) and Total rows (MCP & API): revoking stops new usage but doesn't give back usage that has already been counted.

Reading Your Usage Stats

The Usage This Billing Cycle panel at the top of the page shows your organization's activity for the current billing cycle:
  • MCP requests: The number of requests your active MCP connections have made this billing cycle.
  • MCP rows: The number of rows of data returned to your active MCP connections this billing cycle.
  • Connection status: Active this month once your organization has used the MCP server this billing cycle, or Not yet connected.
  • Total requests (MCP & API): Requests used across MCP and the REST API, shown against your plan's request limit for the billing cycle, with a progress bar.
  • Total rows (MCP & API): Rows used across MCP and the REST API, shown against your plan's row limit for the billing cycle, with a progress bar.
The two Total figures are the ones that count toward your plan's shared MCP and API limits, so they are worth checking periodically. These stats reflect only your organization's own activity; any troubleshooting activity by the PMA support team is excluded and never counts toward your totals. See Get Started with Power My Analytics API for the current per-plan limits. The REST API page shows the same panel, with API requests and API rows in place of the MCP figures.
Notes
The Tools section of your Hub's Overview page shows MCP Calls and MCP Rows as all-time totals, not billing-cycle totals, so those numbers are usually higher than the ones on the MCP page. Use the MCP page when you want to check usage against your plan limits.

What Counts as a Request and a Row

  • Row: One row of data returned from a data source report when your AI assistant (or an API integration) asks PMA for data. Each row in the response counts toward your row allowance for the billing cycle. Only successful responses count; a request that fails adds no rows.
  • Request: Each message your AI assistant sends to the MCP server. That includes the messages it sends when it connects and looks up the available tools, not only your questions, so your request count is usually higher than the number of questions you asked. Requests that don't return report data (such as listing your data sources) count as requests but add no rows.
For example, "What did I spend on Facebook Ads last month?" returns about one row, while "Show me daily spend for every campaign over the last 90 days" for 50 campaigns returns about 4,500 rows. Both are a single question, which is why rows can climb much faster than requests.

Finding Which Connection Is Using Your Requests and Rows

If your usage is climbing faster than expected, the Requests and Rows columns show where it's coming from.
  1. Open the MCP page (MCP & API > MCP).
  2. In the Connected application table, click the Rows column header to sort by rows. Click it again to put the highest number at the top.
  3. Hover over the Requests or Rows header to see which billing cycle the numbers cover (for example, "Usage for billing cycle Sep 24 – Oct 23, 2026").
  4. Check Connected by and Last active on the top rows to see who owns each connection and when it was last used.
  5. To check API keys, open MCP & API > REST API. The Organization API keys table has the same Requests and Rows columns for each key.
What you'll see: one row per connection or API key, with the requests and rows it has used in the current billing cycle. Anything with no usage this billing cycle shows 0.
A few things to keep in mind:
  • One person can appear more than once. Each AI client a person connects (for example, Claude Desktop and ChatGPT) is its own connection. Add up that person's rows to get their total.
  • Rows can grow much faster than requests. A single request can return one row or thousands, depending on the date range and how much detail is asked for. A connection with only a few requests can still use most of your rows.
  • There's no breakdown by data source. The table shows usage per connection, not per connector.
  • The table may add up to less than your totals. Requests and rows from connections you've revoked still count in Total requests (MCP & API) and Total rows (MCP & API), but they no longer appear in the table.
To bring usage down, talk with the owner of the busiest connection about the questions they're asking their AI assistant. Shorter date ranges and summary totals use far fewer rows than daily, campaign-level detail over long periods.

When You Reach Your Limit

  • Per-minute limit: If your AI assistant sends requests faster than your plan's per-minute limit, it gets a "too many requests" error. Wait about a minute and try again.
  • Billing-cycle limit: Once Total requests (MCP & API) or Total rows (MCP & API) reaches your plan's limit, new MCP and API requests are blocked with a "Monthly API/MCP quota exceeded" message until your next billing cycle starts. Your AI assistant may show this as a "too many requests" (429) error. To keep going sooner, buy a Capacity Pack or turn on Premium Unlock from Billing > Add-ons > MCP & API > Increase Limits (see Increase Your MCP and API Limits with Capacity Packs or Premium Unlock), or upgrade your plan; on the Custom plan, contact your account manager.
  • Your reports keep working: Data Studio, Google Sheets, Microsoft Excel, and Power BI reports aren't blocked when you reach your MCP and API limits.
Still stuck? Submit a ticket with your Organization ID and the connection you're asking about.