Destinations
Data sources
AI insights
About us
Careers
Security
Customer reviews
Contact us
Affiliate program
Solution partners
Template gallery
Looker Studio templates
Power BI templates
Excel templates
Google Sheets templates
Tableau templates
Facebook Ads templates
Google Ads templates
Data fields & Metrics
AI prompt library & Guides
Product documentation
API documentation
Case studies
Blog
Data models
Windsor vs Supermetrics
Windsor vs Funnel.io
Windsor vs Fivetran
Windsor vs Rivery
Windsor vs Portermetrics
Windsor vs Coupler.io
Last updated: 14 October 2026 
Windsor MCP is Windsor.ai’s MCP server: one connector that gives Claude, ChatGPT, Copilot, Gemini, Perplexity and Cursor access to every data source you connect to Windsor.ai, from Google Ads, Meta Ads and GA4 to Shopify, HubSpot and 350+ more. You ask a question in the assistant’s chat, and it pulls the data it needs through Windsor.ai.
On Google Ads, Meta Ads and the other platforms that support write actions, the assistant can also make changes, such as pausing a campaign or adjusting a budget, after you confirm each one.
🚀 Connect your first data source to Windsor.ai for free, then add Windsor MCP to your assistant: https://onboard.windsor.ai/.
What is Windsor MCP?
MCP (Model Context Protocol) is an open standard, first published by Anthropic, that lets AI assistants call external tools through a defined interface. Windsor MCP is Windsor.ai’s implementation: it gives the assistant tools to list your connected sources and accounts, look up the valid field names for each source, and fetch data for the dates you ask about.
Because one connector covers every source, a single question can combine platforms, for example Google Ads spend with GA4 key events, or Meta Ads spend with Shopify orders. Spend, clicks and impressions have the same names on every ad platform, so the assistant can add them up without renaming columns.
What it looks like with real data
We asked Windsor MCP for sessions and key events by channel group from windsor.ai’s own GA4 property, for the 30 days from 8 September to 7 October 2026. The query used three GA4 fields: session_default_channel_group, sessions and conversions (key events). The result, shown as rates:
| Channel group | Share of sessions | Key events per session |
|---|---|---|
| Direct | 31.7% | 18.6% |
| Referral | 23.9% | 15.8% |
| AI Assistant | 21.1% | 24.2% |
| Organic Search | 15.7% | 7.4% |
| Paid Search | 1.2% | 6.9% |
Visits from AI assistants made up a fifth of our sessions in that period and recorded more key events per session than direct, referral or search visits. A follow-up question in the same chat, such as which landing pages those visits arrived on, takes one more prompt.
What teams use Windsor MCP for
- Cross-channel reporting: “Compare spend, clicks and cost per conversion across Google Ads, Meta Ads and TikTok Ads for last month, one row per platform.”
- Client reporting for agencies: “For [client], give me last month’s results across Google Ads, Meta Ads and GA4 against the month before, with three takeaways.”
- Diagnosis: “Paid search conversions fell last week. Which campaigns, and did landing page sessions in GA4 fall at the same time?”
- Ecommerce: “What was ROAS by channel last month, using Shopify revenue and ad spend from every platform?”
- Actions: “Which Google Ads search terms spent over $100 without a conversion? Add them as negative keywords after I confirm.”
For 30+ prompts by role, see Windsor MCP examples and use cases, and for paid media, Windsor MCP for PPC.
See Windsor MCP in action:
How to set up Windsor MCP
Connect at least one data source in your Windsor.ai account first. You do not need to select fields or build a query in Windsor.ai for the assistant: every connected account and field is available to it. Then add Windsor MCP to the assistant you use.
Windsor MCP for Claude
Windsor.ai has a native connector in Claude: open the Connectors menu, search Windsor.ai and click Add. Follow the Claude setup guide for the full steps.
Windsor MCP for ChatGPT
Windsor.ai is available as a ChatGPT plugin: click Plugins in the ChatGPT sidebar, search Windsor.ai and click Install plugin. Follow the ChatGPT setup guide for the full steps.
Windsor MCP for Copilot Agent
You need an existing agent in Copilot Studio. Follow the Copilot Agent setup guide.
Windsor MCP for Perplexity
You need the Perplexity desktop app. Follow the Perplexity setup guide.
Windsor MCP for Cursor
You need the Cursor desktop app. Follow the Cursor setup guide.
Windsor MCP for Gemini
You need the Gemini CLI, which requires Node.js 18 or later. Follow the Gemini setup guide.
Conclusion
Windsor MCP gives your AI assistant one connection to all your marketing and business data, with the same field names across ad platforms and write actions on the platforms that support them. Connect the two or three sources you report on most, add Windsor MCP to Claude or ChatGPT, and start with a cross-channel question you currently answer in a spreadsheet.
In a ChatGPT Business, Enterprise or Edu workspace, Windsor MCP can also run inside a workspace agent that sends a report on a schedule. ChatGPT workspace agents with Windsor.ai has the setup, with a weekly Google Ads report as the example.
🚀 Get started with Windsor.ai on the free forever plan, no credit card: https://onboard.windsor.ai/.
FAQs
Do I need a paid Claude or ChatGPT plan?
No for Claude: Windsor.ai can be added as a connector on Claude’s free plan, though free usage limits run out sooner on long analyses across several platforms, and teams that use it daily usually move to a paid plan. On ChatGPT, which plans can install plugins depends on your ChatGPT plan and workspace settings.
Does the assistant see all my data sources, or only one?
All of them. One Windsor MCP connection covers every source and account connected to your Windsor.ai account, so the assistant can combine Google Ads, Meta Ads, GA4 or Shopify in one answer.
Do I need to select fields in Windsor.ai before asking?
No. The field selection in the Windsor.ai app is for exports to sheets, BI tools and warehouses. The assistant can use every field and account you have connected, and it looks up the right field names itself.
Can I use Claude and ChatGPT with the same Windsor.ai account?
Yes. Add Windsor MCP to each assistant and sign in with the same Windsor.ai account. Team members invited to that account can do the same, and every paid plan has unlimited users.
How does the assistant know which client's account I mean?
Name the client or the account ID in your prompt. In Claude, a project per client with that client’s account IDs in the project instructions keeps every chat on the right accounts.
Can Windsor MCP change my campaigns?
Yes, on platforms that support write actions, including Google Ads and Meta Ads. The assistant shows what it is about to change and waits for your confirmation before each action, and write actions can be switched off if you want read-only access.
Why use Windsor MCP instead of each platform's own MCP server?
Platform MCP servers, such as Google’s for Google Ads, cover one platform each. Windsor MCP covers 350+ sources through one connection, so one question can compare platforms and blend them with GA4 or CRM data. See Windsor.ai vs the Meta Ads MCP for a detailed comparison.
Does Windsor.ai store my data?
Windsor.ai fetches data from each platform’s API when the assistant asks and keeps a short-term cache so repeated questions are faster. For ad and analytics platforms that means aggregated metrics such as spend, clicks and sessions; CRM sources can also include contact records. See security for details.
