Model Context Protocol

Connect Claude, ChatGPT & any AI tool to your store data

Meta says one number, Google another, Shopify a third, and the AI you would like to ask sits in a fourth tab. Datadrew's MCP server gives it a direct, read-only line to your Shopify store and your Meta and Google ad accounts. Ask in plain English and get blended ROAS, cohort LTV and product-level results back in seconds. Real numbers from your connected data, not guesses. Set up in 2 minutes.

Start for free See the 2-minute setup

Works with

ClaudeClaude ChatGPTChatGPT CursorCursor WindsurfWindsurf Visual Studio CodeVS Code Any MCP client
Claude Claude
Datadrew connected
Why did our blended ROAS drop last week?
D
Called Datadrew · fb_ads_get_campaign_insights
Blended ROAS2.1x -18%
Meta CPC$1.74 +24%
MER4.1x -9%
Meta CPC jumped 24% on 3 prospecting campaigns while Google held steady. Shift ~15% of budget to Google Brand and pause the 2 fatiguing ad sets.
Ask your store anything…
1,000+Shopify brands
$1.5Bn+GMV analyzed
15+Data sources, one server
47+Countries

Three tabs, three revenue numbers, and an AI that has seen none of them

Meta says one number, Google another, Shopify a third. The AI you would like to ask about it sits in a fourth tab, working from whatever you pasted in last.

Confident answers with no data behind them

"LLMs can sound brilliant and still be wrong. In live accounts with meaningful spend, confident mistakes are expensive." That is Ali Mehdi Mukadam on PPCChat, and he is right. Ask about last week's ROAS with nothing attached and you get a plausible figure and a plausible reason. Neither came from your account.

Platform claims that add up to more than your orders

How operators put it to each other: "added up what meta and google each claim they drove. it's more conversions than orders i actually got." Each platform self-attributes. Summed, the claims exceed the orders Shopify recorded, and "which one gets next week's budget" has no answer inside either ad account.

Every Monday, the same export

Shopify, Meta and Google exported by hand, stitched in a sheet, pasted into a chat window. Then again next week, because last week's paste is last week's data. That is a lot of spreadsheet gymnastics for one question: did the ads make money?

Datadrew MCP puts your Shopify orders and your Meta and Google ad accounts behind one read-only connector, so the AI you already use reasons on your numbers instead of guessing at them.

Your ad accounts, not just your store

Which MCP gives Claude or ChatGPT my Shopify orders and my Meta and Google ad spend through one connector?

Datadrew MCP. One endpoint, one OAuth sign-in, and the AI client you already use reads Shopify orders, Meta Ads and Google Ads together, with the same metric definitions the Datadrew app uses. Read-only, and free to connect on every plan.

Google's official server reads Google Ads; Meta's reads and writes Meta Ads. The questions that decide budget live in neither: "Is Google's 2.2x better than Meta's 1.9x?" cannot be settled inside either account, because the two figures use different attribution windows and both claim some of the same orders. What Shopify orders give you is the check: whether the two claims together exceed what the store actually recorded, and whether total spend made money at all. That is where the Google Ads MCP guide starts: connect Google Ads and Meta Ads in Datadrew, paste the endpoint into Claude or ChatGPT, and ask for yesterday's spend on both against yesterday's orders. A Meta Ads MCP guide in the same shape is next.

Start with what you're connecting

"Which MCP server does my Shopify store actually need?" Shopify's own servers, Datadrew's, and when each is the right pick. Read the guide → "I want Shopify in Claude, and a dashboard I can share" The connect steps, the build-it-yourself route, and a dashboard Claude refreshes through MCP. Connect Shopify to Claude → "I want Google Ads and my store in Claude or ChatGPT" Google's official read-only server, Datadrew's, and a 10-prompt pack that sets Meta and Google spend against Shopify orders. Google Ads MCP guide → "I run operations on Unicommerce in India" Warehouse, returns and delivered-versus-ordered data next to your Shopify orders, inside Claude. Unicommerce in Claude →
"I want my Meta ad account in Claude" Meta's official server, Datadrew's read-only one, and the creative prompts that need Shopify in the room. For now: Meta creative analysis in Claude. Coming
"I want the Monday marketing report written for me" A scheduled Claude or ChatGPT report from Datadrew MCP, in your inbox before you are at your desk. Coming
"I run an agency and need every client store in one chat" One connector, every client, isolation you can explain to each of them. The multi-store block below has the basics today. Coming
"I use ChatGPT, not Claude" The ChatGPT setup, plan by plan, with your Shopify data. Until it lands, the ChatGPT tab below has the steps. Coming
How it works

One connector. All your data. Plain English.

MCP (Model Context Protocol) is an open standard that lets AI assistants securely connect to outside tools and data. Datadrew runs a public, OAuth-secured MCP server at mcp.datadrew.io. Connect it once and your AI client can query every platform you've linked to Datadrew, answering with the exact same numbers, metric definitions and permissions as the app.

1

Open Datadrew MCP

In Datadrew, go to Settings → Datadrew MCP. It's free to connect on every plan, including Free.

2

Copy your secure URL

Copy the Datadrew MCP URL, mcp.datadrew.io/mcp, shown right here and in Settings → Datadrew MCP.

3

Paste into your AI tool

Add it as a connector in Claude, ChatGPT, Cursor or any MCP client, authorize with OAuth, and start asking.

Server endpoint https://mcp.datadrew.io/mcp

Paste this URL into your AI tool, then authorize with OAuth. You'll also find it, with step-by-step instructions, under Settings → Datadrew MCP inside Datadrew.

Set it up in your favourite AI tool

Pick your client and follow the steps. Same URL, same OAuth flow, everywhere.

Your MCP URL https://mcp.datadrew.io/mcp
  1. Open Settings → Connectors in Claude (labelled Customize → Connectors on some accounts; works in Claude.ai and Claude Desktop).
  2. Click Add custom connector and paste the URL above.
  3. Click Connect and authorize with your Datadrew account via OAuth. If you manage more than one store, pick the one this chat should use.
  4. Start any chat and ask "What was my blended ROAS last week?" Drew's tools run automatically.

Custom connectors are available on Claude Free, Pro, Max, Team and Enterprise; a Free account gets one custom connector. Source: Anthropic's custom-connector guide, checked 10 September 2026.

  1. In ChatGPT on the web, turn on developer mode in Settings, then open Connectors.
  2. Add a custom connector and paste the URL above.
  3. Sign in with OAuth to link your Datadrew workspace and pick the store.
  4. Enable the Datadrew connector in the chat's tool menu and ask. ChatGPT calls Datadrew for the real numbers.

ChatGPT Business, Enterprise and Edu workspaces get full custom-connector support. Pro (and Plus, per OpenAI's developer docs) can add read-only connectors in developer mode, which is all this server needs. Web only; Free is not listed. Source: OpenAI's developer mode and MCP apps page, checked 10 September 2026.

  1. Open your MCP config, mcp.json, in Cursor, Windsurf or VS Code.
  2. Add a server entry pointing at https://mcp.datadrew.io/mcp.
  3. Reload and complete the OAuth flow when your editor prompts you.
  4. Drew's read-only tools appear alongside your other MCP servers, ready to query.

Claude Code is one line in the terminal; the first tool call opens a browser for sign-in.

claude mcp add --transport http datadrew https://mcp.datadrew.io/mcp
  1. Point any spec-compliant MCP client at https://mcp.datadrew.io/mcp (streamable HTTP).
  2. Authenticate with OAuth. Access is scoped to your workspace and plan.
  3. Discover Datadrew's read-only tools across every connected platform.
  4. Use them from any agent framework, automation or custom app that speaks MCP.
The MCP server is free to connect on every plan. You'll find the URL and full instructions any time in Settings → Datadrew MCP.
15+ sources, one connector

Everything your AI can pull

Connect once and your assistant can query 15+ platforms in plain English, all blended into consistent CAC, ROAS, MER and LTV definitions the AI actually understands.

Store & orders
ShopifyShopifyQLGraphQLInventory

"Show me yesterday's revenue, AOV and units sold by product type."

Advertising
Meta AdsGoogle AdsAmazon Ads

"Compare ROAS and CPC across Meta and Google for the last 14 days."

Web & SEO
Google Analytics 4Search Console

"Which landing pages convert best, and which search queries are rising?"

Email & retention
KlaviyoBrevo

"How much revenue did my Klaviyo flows drive vs campaigns this month?"

Payments & subscriptions
StripeRechargeSkio

"What's my subscription churn and MRR trend by plan?"

Marketplaces & ops
Amazon SellerUnicommerceJudge.meAfterShip

"Break down sales by marketplace and flag any late shipments."

What's behind the socket

Why the join matters: the month both platforms claimed the same orders

The connector is the easy part. What sits behind it is Shopify orders as the number every platform claim gets checked against.

"So they make budget decisions based on bad data. They over-invest in whatever platform tells the prettiest story."Curtis Howland, DTC growth operator (source, verified)
Worked example · illustrative numbers

A brand doing about $1.7M a year, 30 days, Meta and Google both running

Each platform reports conversions in its own attribution window. Shopify records the orders that happened. Side by side:

What each system reportsMeta AdsGoogle AdsShopify orders (+ total spend)
Spend, 30 days$62,000$34,000$96,000, both platforms summed
Conversions it reports1,180 purchases760 conversions1,420 orders, all channels
Revenue it reports, at a $98 AOV$115,640$74,480$139,160
Return on its own scorecard1.87x ROAS2.19x ROAS1.45x MER (all revenue ÷ all spend)

Summed, the platforms claim 1,940 conversions. The store recorded 1,420 orders, some of them from email and direct. So at least 520 orders were claimed twice. That is arithmetic, not a verdict on which platform deserves them; the table cannot tell you that, and nothing on this page claims to.

What it does settle: the business made 1.45x on total spend. At a 60% margin (the margin you know from your own books; no connector computes it) that is $83,496 of gross margin against $96,000 of ads, about $12,500 lost on advertising before a single fixed cost, while two dashboards reported healthy ROAS. Blended MER is an efficiency ratio that credits every platform with all revenue. It will not tell you which campaign to cut; it is the first number that says whether there is anything to cut at all.

In one documented account, Meta and Google each self-attributed the same purchases. Roughly £4,200 a month was misallocated for five months, about £21,000, before anyone summed the platform claims against Shopify orders. The full story, and the prompts that run this check on your own data, are in the Google Ads MCP guide. The method behind the ratio: blended ROAS across Meta, Google and Shopify and which ROAS number you can trust.

Six things a per-platform connector cannot hand your AI

Even a well-kept build-it-yourself connector returns raw platform data. Six things stay missing, and the scope table further down carries the detail:

  • Blended math against Shopify ordersMER, blended CAC and blended ROAS from all spend and Shopify revenue, the double-count sized as above.
  • One definitions spineCAC, ROAS, MER and LTV defined once, so Claude's answer matches the app and last month's answer.
  • Daily warehouse historyYour Shopify, Meta and Google data warehoused day by day: three months of history on Free, all-time on AI Intelligence and AI Ads CoPilot. Daily, not real-time: a weekend anomaly surfaces Monday.
  • A loadable analysis methoddatadrew_load_skill hands the model Datadrew's analyst notes for an ad-account audit, a wasted-spend review or a "what changed" investigation.
  • Read-only by designEvery tool reads. It can never pause campaigns, edit orders, move budget or place spend. "You set the strategy. It writes the code." is how Ali Mehdi Mukadam splits the work between operator and LLM in live ad accounts. The server side of that split is read-only.
  • Per-store isolationOne connector, many stores, each answer scoped to one shop and labelled with it.

Run this month's numbers on your own store

Install Datadrew free on Shopify, connect Meta and Google, paste the endpoint into Claude or ChatGPT and ask for last month's spend on both against Shopify orders. This check runs on the read tools, which work on Free. The computed analyst tools (cohort LTV, creative intelligence) need AI Intelligence or AI Ads CoPilot (plans).

Install free on Shopify Paste the endpoint

A prompt library to get you started

Copy a prompt, paste it into Claude or ChatGPT, and let Drew pull the numbers. Filter by what you're working on.

The budget prompts only work when ads and store share one socket. "Where should I shift budget this week" needs Meta spend, Google spend and Shopify orders in one answer; a single-platform connector returns one of the three. The retention prompts run on Shopify alone, and so do most of the product prompts. The ads-specific set, a 10-prompt pack naming the tool each prompt runs (google_ads_run_gaql to datadrew_cohort_ltv), lives in the Google Ads MCP guide.

Acquisition

Where should I shift budget this week for better blended ROAS?

Acquisition

Show me which ads are winning and which to pause across Meta and Google.

Acquisition

Which campaigns wasted the most budget last week, and why?

Retention

Show me my customer segments: who are my champions vs at-risk?

Retention

What's my 90-day LTV by first-order month?

Retention

Which products drive the most repeat purchases?

Product

Which SKUs are hero products vs ad-spend wasters?

Product

What are my top product bundles by revenue for cross-sell?

Product

Show me repurchase rates by product type over the last 6 months.

Executive

Give me an executive summary of this month's performance.

Executive

How are we performing vs last month, and what changed?

Executive

Which channel should I scale next, and what's the expected payback?

For agencies & multi-store brands

Every client store, in one conversation

Datadrew MCP covers every Shopify store your login manages. No re-authing between clients, no juggling logins. Ask, and name the store.

  • Query multiple stores in a single chat session.
  • Name the store you want, or ask for two stores back to back; the connector works one shop at a time and switches when you ask.
  • Every answer reports which store it actually ran against.
  • One Datadrew login, OAuth-scoped to exactly the shops you manage.

Isolation is the property you can explain to a client: the connector is scoped to one shop at a time, every answer says which shop it ran against, and one client's data never appears in another client's answer. Switch with "List my Datadrew shops" and "Switch Datadrew to [shop name]", with no re-authorising. An agency setup guide is next; until then, live client dashboards from Claude covers the reporting half.

AAcme Cosmetics● connected
NNorthwind Apparel● connected
PPeak Supplements● connected
"Switch Datadrew to Peak Supplements and show 30-day MER →"
Secure by design

Your data, on your terms

Plug AI into your store with confidence. Datadrew MCP is read-only, OAuth-secured, and built so your assistant can analyze everything while changing nothing.

OAuth, not API keys

Access is granted through OAuth and tied to your Datadrew workspace. Revoke it anytime from the app. No secret keys to leak.

Read-only by design

The server can read and analyze your data, and it can never pause campaigns, edit orders, move budget or place spend.

Data minimization

Drew sends your AI only the data needed to answer the question in front of it. Nothing more, never your whole database.

Encrypted & compliant

Data is encrypted in transit and at rest on SOC 2-compliant, GDPR-aligned infrastructure, the same backbone that powers the app.

Datadrew MCP, a connector you build, or a platform's official MCP: what each one returns

Three options. The right one depends on whether your question lives inside one ad account or across the business.

Scope Datadrew MCP Claude or ChatGPT with a connector you build and maintain An official platform MCP (Google Ads, Meta)
Data covered Shopify orders, products, customers, inventory; Meta Ads; Google Ads; GA4, Klaviyo and 15+ sources, one endpoint Whatever you wire up and keep wired; each source is its own auth and code One platform. Google's reads Google Ads; Meta's reads and writes Meta Ads
Read vs write Read-only by design, OAuth-scoped; cannot pause campaigns, edit budgets or place spend Your call; you own the risk and the guardrails Google: strictly read-only, three tools. Meta: read and write, open beta
History Daily warehouse history: three months on Free, all-time on paid plans The platform API's window, unless you build and pay for storage The platform API's window
Metric definitions One spine: CAC, ROAS, MER and LTV defined once, the same as in the app Whatever the model derives from raw fields, session by session The platform's own; Google ROAS and Meta ROAS are not comparable with each other
Blended math against Shopify orders Yes: MER, blended CAC, and the double-count sized Only if you join it yourself, every time No. The store is out of scope
Analysis method Loadable analyst playbooks (datadrew_load_skill) Your prompts None beyond the tool descriptions
Multi-store isolation One connector, many stores, each answer scoped to one shop and labelled with it One connector per store, or a switcher you build Per platform login; no store-level scoping, because the store is out of scope
Cost to connect Free to connect on every plan; computed analyst tools need AI Intelligence or AI Ads CoPilot (plans) Your engineering time and hosting Google: open source, runs locally or on your own Cloud Run. Meta: through your own Meta app. Neither page states a price as of 10 September 2026

Google and Meta columns from their own developer pages, checked 10 September 2026. Links below.

Building your own is a respected path, and if your question is one platform deep, an official server may be all you need. What neither gives you is the join: Google's ROAS and Meta's are not rankable against each other, so "where does next week's budget go?" cannot be settled inside either account. Against Shopify orders you can at least see whether the two claims add up to more than the store recorded, and whether total spend made money. That check is the reason this server exists.

Other MCP servers you will run into, in their own words

Google Ads MCP Server (open source, announced 7 October 2025). Read-only, three tools: list_accessible_customers, search for GAQL, get_resource_metadata. Google's docs: "This implementation is strictly read-only. It cannot modify bids, pause campaigns, or create new assets." Runs locally over stdio by default; needs a developer token plus OAuth 2.0 or a service account. Google's documentation.

Meta Ads MCP Server (open beta since 29 April 2026; any developer's own Meta app since 16 July 2026). Read and write: "Create and edit ads, ad sets, and campaigns using natural language", plus reporting and catalogs. Meta's announcement.

Triple Whale MCP. Read-only (moby:read) on eligible Triple Whale plans, framed in its knowledge base (read 10 September 2026) as "Triple Whale MCP gives external AI tools access to your data. Moby helps you take action on that data."

Madgicx MCP. Meta only, with 22 read-only and 27 write and delete tools per its academy page (March 2026); plan gating not verified as of 10 September 2026.

Ryze AI. 14 Google Ads tools (8 read, 6 approval-gated write), with Meta Ads, GA4, Search Console, Shopify and other platforms on the same endpoint (150+ tools in total, by its own count), free to connect with a paid tier at $89 a month, per its Google Ads MCP page updated 3 September 2026.

One analyst, everywhere you work

The MCP server is one of three ways to work with Drew. Same numbers, same definitions, wherever your team lives. The socket is the surface; the agent behind it is the product.

In the Datadrew app

Chat with Drew right inside Datadrew, next to your dashboards, on every plan.

Meet Drew AI
AI Intelligence & AI Ads CoPilot

In Slack

Mention @Drew in any channel and get charts and recommendations back in-thread.

Slack integration
Free to connect on every plan

Via MCP (you're here)

Bring Drew's tools into Claude, ChatGPT, Cursor or any MCP client with one URL.

Jump to setup

What Drew does with this today, and what it does not do yet

  • LiveThrough MCP: the tools that read your own connected accounts on every plan; the analyst tools whose answers Datadrew computes (cohort LTV, creative intelligence, competitor tracking) and Drew's analyst playbooks in-session on AI Intelligence or AI Ads CoPilot.
  • LiveIn the app and Slack: Diagnose Performance (why a metric moved, across Shopify, Meta and Google) and Budget Recommendations (what to scale, reduce or pause, how much, and why).
  • PilotThe Daily Ads Brief, a diagnosis-grade daily brief, is running on pilot accounts.
  • Rolling outExecute Ads Changes: approved changes with guardrails, rolling out now. Never through MCP; this server stays read-only.
  • Not builtAn out-of-stock spend detector or a negative-margin SKU sweep. Both are analyses you can run, or ask Drew to run, today.

Connect Datadrew's MCP from Settings → Datadrew MCP and ask Claude or ChatGPT for yesterday's Meta and Google spend against your Shopify revenue, product-level results and cohort LTV: one connector, one set of definitions, read-only by design. The socket is where you work; the judgment is the product. Meet Drew, the AI ads agent for Shopify brands.

Trusted by fast-growing Shopify brands

★★★★★

"One of the best apps in the Shopify App Store. Saves hours of manual reporting if you track customer lifetime value. Good options for slicing and dicing by customer cohort, product, etc."

KH
Kyle HillDirector of Paid Media, KradleMyPet
★★★★★

"WOW is all I can say. I have spent so much time trying to nail down an accurate LTV and I was amazed at how robust this app is and gives me quality data that I can take action on."

EB
Eric BirkemeierCo-Founder & CMO, Shred Lights
★★★★★

"Needed a tool for historic cohort / LTV analysis. Datadrew hit the sweet spot. Nifty, to the point, quick 1:1 support and great pricing."

SF
Stephan FrehPartner

Frequently asked questions

MCP, the Model Context Protocol, is an open standard that lets AI assistants like Claude and ChatGPT securely connect to external tools and data. Datadrew runs a public, OAuth-secured MCP server at mcp.datadrew.io, so your AI client can query the platforms you've connected to Datadrew using your own data instead of guessing.
Any MCP-compatible client, including Claude (Claude.ai and Claude Desktop), ChatGPT, Cursor, Windsurf, VS Code and other agent frameworks that support the Model Context Protocol. You connect them all with the same MCP URL and OAuth flow.
Yes. Access is granted through OAuth and tied to your Datadrew workspace, so you can revoke it anytime. The MCP server is read-only, it sends your AI only the data needed to answer the current question, and all data is encrypted in transit and at rest on SOC 2-compliant, GDPR-aligned infrastructure.
No. The Datadrew MCP server is read-only by design. Your AI assistant can read and analyze your data, but it can never pause campaigns, edit orders, move budget or place spend. It analyzes everything and changes nothing.
Datadrew MCP is free to connect on every plan, including Free. The tools that read your own connected accounts (Shopify, Meta Ads, Google Ads, GA4, Klaviyo and more) work on any plan; the analyst tools whose answers Datadrew computes (cohort LTV, creative intelligence, competitor tracking) need AI Intelligence or AI Ads CoPilot. Your personal MCP URL appears in Settings → Datadrew MCP, and plan details are on the pricing page.
About two minutes. Open Settings → Datadrew MCP in Datadrew, copy your MCP URL, paste it into Claude, ChatGPT or your IDE as a custom connector, and authorize with OAuth. There's no code to write.
Everything you've connected to Datadrew: 15+ sources including Shopify, Meta Ads, Google Ads, GA4, Google Search Console, Klaviyo, Amazon Seller and Amazon Ads, Unicommerce, Stripe, Recharge, Skio, Brevo, Judge.me and AfterShip, plus blended cross-channel KPIs like ROAS, MER, CAC and LTV.
Yes. Multi-store merchants and agencies can work across several stores in a single session: the connector is scoped to one shop at a time, and you switch by saying "Switch Datadrew to [shop name]", without re-authenticating. Every answer tells the AI which store it ran against, and access is OAuth-scoped to exactly the shops you manage.
Yes. Connect Google Ads in Datadrew, add https://mcp.datadrew.io/mcp as a custom connector in Claude, and the google_ads_run_gaql and google_ads_list_accounts tools read your account in the same chat as your Shopify orders. Google also publishes an official open-source Google Ads MCP server: strictly read-only, three tools, runs locally with a developer token. The step-by-step setup and a 10-prompt pack are in the Google Ads MCP guide.
No. Datadrew MCP is read-only by design and OAuth-scoped: it cannot pause campaigns, edit budgets or place spend, whatever you type into the chat. If you want a change made, make it in Ads Manager or Google Ads. Execute Ads Changes, approved changes with guardrails, is rolling out in the Datadrew app and will never run through this server. Some other MCP servers (Meta's official server, Madgicx, Ryze) offer write tools; that is their design choice, and this server's choice is to read.
Yes. One connector covers every store your Datadrew login manages, scoped to one shop at a time. Say "List my Datadrew shops" or "Switch Datadrew to [shop name]" in the chat to move between clients without re-authorising. Every answer names the shop it ran against and never mixes two clients' data, which is the property you can explain to a client. An agency setup guide is coming.

Bring your store data into Claude & ChatGPT

Connect Datadrew's MCP server and ask your AI anything about your Shopify orders and your Meta and Google ad accounts. Securely, with OAuth, in two minutes.

Install free on Shopify Book a demo

Free to connect on every plan. Analyst tools need AI Intelligence or AI Ads CoPilot.