Datapipe
Connect the sources once
One project can hold the tools you already use. You do not install a new plugin for each login.
Compare · ChatGPT, Claude, and GrokDatapipe works with your AI,
Datapipe works with your AI,
not instead of it.
Anyone can install a plugin and ask Shopify or HubSpot a question in chat. That is useful. Datapipe is different: you connect your sources once, keep reports refreshing, and send owner-approved changes back to supported sources (Shopify prices, HubSpot records, daily ad budgets) without handing over the logins.
Works alongside your AI · Owner-approved changes · Sheets that refresh on their own
Separate logins in the chat on the left. On the right, Datapipe reads data into reports and sends approved changes back to your sources.
Plugins vs Datapipe
What actually changes
MCP is a shared way for an AI assistant to plug into another app and ask it for information. ChatGPT, Claude, and Grok all let you add plugins or MCP connectors. Datapipe uses that same door so your AI can read a project you already connected, and owner-approved changes go back to supported sources from the same workspace.
| Topic | Plugins in ChatGPT, Claude, or Grok | Datapipe |
|---|---|---|
| What you connect | Each tool separately, with that tool’s login. | Many sources once, inside a project. The data sources page shows what is available now and what is still coming soon. |
| Who holds the login | Usually you, inside the AI app. | Whoever connected the source in Datapipe. Teammates and AI never see that password. |
| When the numbers update | When someone asks in chat. | On a schedule: daily on Free and Starter, hourly on Pro, every 15 minutes on up to 50 Agency sources. Destinations are Sheets, Drive, and BigQuery (Pro and Agency). Data Studio reads the Sheet. |
| Who else can use them | Whoever is in that conversation, with that person’s access. | People you invite to the project, plus an AI you connect to that same project. |
| Can it change a record? | Some official plugins can write. For example, HubSpot’s ChatGPT connector can create and update records, and Shopify’s connectors can change products and collections. | Yes, on supported sources. A teammate or AI proposes a change. The owner reviews the exact current and proposed values, then Datapipe applies it and keeps a record in Actions & approvals. Examples include Shopify prices, HubSpot contacts and deals, and daily budgets on Amazon Ads, X Ads, and StackAdapt. Writes start switched off until the owner turns specific actions on. |
| Where the answer lives | In the chat thread. | In shared reports and destinations. You can still ask Claude or ChatGPT about the same project. |
| Cost | Official plugins are often free beyond the AI plan and the tool you already pay for. | Free for 3 data sources. Starter is $15 for 25, Pro $49 for 100 with hourly refresh and BigQuery, Agency $199 for 500 with 15-minute refresh on up to 50 of them. Unlimited projects on every plan. |
AI app support and plan requirements vary. ChatGPT, Claude, and Grok each decide which plugins you can add and what those plugins are allowed to do.
When a plugin is enough, and when it is not
A plugin is a good fit for one person asking a one-off question about one tool. Datapipe is a better fit when several people need the same numbers, a Sheet should refresh without a chat, or you want owner-approved changes sent back to Shopify, HubSpot, or supported ad accounts, without each login inside the AI app.
Your AIShopifyHubSpotAnalyticsStripe
Meta AdsYour AI
SheetWhat Datapipe adds around the chat
| Metric | Today |
|---|---|
| Sessions | 12,840 |
| Orders | 318 |
| Revenue | 18,420 |
DailyHourly15 min
The Sheet updates without a chat
Daily, hourly, or every 15 minutes, by plan, even if nobody asks.
Project addressYour AI reads a project, not every login
Paste the project address into Claude or ChatGPT so it can read that project’s reports. The token cannot see other projects.
Shopify · Weekender bag
Propose a change. The owner reviews it.
On supported sources (Shopify prices, HubSpot records, and daily budgets on Amazon Ads, X Ads, or StackAdapt), a teammate or AI proposes the change. The owner reviews the exact values, then Datapipe applies it and keeps a record.
Connect your sources. Read, report, and change, with you in control.
Start free with 3 data sources. Keep reports refreshing, and turn on owner-approved writes when you are ready.
Cancel anytime
Questions before you start
Plugins, MCP, and Datapipe
Plain answers about what each side can do.
What is MCP, in one sentence?
MCP is a shared way for an AI assistant to plug into another app and ask it for information. Datapipe’s connector uses that door so Claude or ChatGPT can read reports in one project.
Can’t I just install the Shopify or HubSpot plugin?
Yes, if you are one person asking a one-off question about that one tool. Official plugins exist, including HubSpot for ChatGPT and Shopify’s connectors for ChatGPT, Claude, and Grok. Use them when that is all you need. Add Datapipe when several people need the same numbers, when a Sheet should refresh without a chat, when you want owner-approved changes back to supported sources, or when you do not want to put each login into the AI app.
Does Datapipe replace ChatGPT, Claude, or Grok?
No. Datapipe is the data layer those assistants can connect to. You still ask questions in your AI. The difference is that the assistant reads a Datapipe project instead of holding a separate login for every tool. When a change is needed on a supported source, the assistant can propose it in Datapipe; only the owner approves.
Does this work with Grok?
Datapipe’s connect screen is built for Claude and ChatGPT. The in-app help guide also covers Cursor. Grok can add custom MCP servers, according to xAI’s connector docs, but we have not verified Datapipe’s connector with Grok, so we do not claim that it works there yet.
Can the AI change my store, CRM, or ads?
Yes, on supported sources, with you in control. A teammate or AI proposes a change through Datapipe. The workspace owner reviews the exact current and proposed values, then Datapipe applies it and keeps a record in Actions & approvals. Writes start switched off until the owner turns specific actions on. Examples include changing a Shopify price, updating a HubSpot contact or deal, and changing a daily budget on Amazon Ads, X Ads, or StackAdapt. Datapipe’s MCP connector (the project address you paste into Claude or ChatGPT) is read-only. It can query that project’s reports and cannot change records or see source passwords. AI assistants propose writes through Actions & approvals, the same way a teammate can, and they cannot approve them. Some official plugins can write on their own. HubSpot’s ChatGPT connector can create and update records, and Shopify’s connectors can change products and collections, without Datapipe’s owner-review step.
Will Datapipe remember history that a chat does not?
A plugin answers in the conversation. Datapipe’s scheduled reports land in Sheets, Drive, or BigQuery, so the same table is there tomorrow. New projects use Live data by default and do not keep report rows inside Datapipe. You can turn on Saved history for a project when you want Datapipe to keep those rows (90 days at no charge, longer with Data Vault).
What if I need a source Datapipe does not have yet?
Then a dedicated plugin or MCP may be the right next step. For example, Google Ads is listed as coming soon on Datapipe’s data sources page. We do not claim it works in Datapipe today. That page marks what is available now and what is still coming soon.
How much does Datapipe cost next to a free plugin?
Official plugins are often free beyond the AI plan and the tool you already pay for. Datapipe is free for 3 data sources, then $15 a month for 25 (Starter), $49 for 100 with hourly refresh and BigQuery (Pro), or $199 for 500 with 15-minute refresh on up to 50 of them (Agency). You are paying for scheduled reports, team access, project isolation, and owner-approved writes, not for the chat itself.
Still unsure? Contact us.