MCP, or Model Context Protocol, is an open standard from Anthropic that solves one simple problem: how to give an AI direct access to external data and tools so it can understand on its own what to do with them.
We’ve integrated this protocol into MTWSPY so you can analyze search results using AI in 2-5 minutes, with just a single request.
Now you can query the ad database, apply filters by geo, vertical, date, and volume, and aggregate data on the fly — whether that’s top offers for a specific country or vertical trends over the past month.
No more manually sorting through cards and exporting to Excel: you phrase your request in plain language, and the AI handles all the routine filtering and calculations.
In this article, we’ll show you how to set up MCP in MTWSPY, how views are deducted, and walk through a few usage examples.
Important: treat your API key like your account password — don’t publish it or share it openly.
Setting Up MCP in Claude + MTWSPY
To get started, you’ll need:
- an MTWSPY account with a Basic subscription or higher
- an AI that supports MCP, such as Claude, ChatGPT, Gemini, etc.
We’ll walk through the setup using Claude as an example.
Step 1. Sign up for MTWSPY and purchase a subscription to get your API key.
If this is your first purchase, use promo code WELCOME for 25% off. For example, on the Basic plan (30,000 views/month), that’s $52.5 instead of $70.
Step 2. Click your email in the bottom left corner and select “Account Settings”:
Go to the “API” section, click “Create Key,” and give your key a name.
Step 3. Copy your new API key, open Claude, and select “Settings”
Step 4. In the window that opens, select “Connectors” and click the “+Add” button. Then click “Add custom connector”:
Step 5. Enter a name, e.g. MTWSPY, and enter the MCP server URL — https://mcp.mtwspy.com/mcp. Click “Continue”:
Step 6. In the window that appears, scroll down to the “Request headers” section, select x-api-key, and paste in the key you got in step one:
Step 7. Final step — click “Add” and “Connect,” and you’ll see everything is connected:
What You Can Do with MCP + MTWSPY
Now ad analysis takes less time and can be delegated. The AI draws on the MTWSPY database and will answer any question you have using that data.
The AI doesn’t “make up” data — it pulls it from MTWSPY. For narrow queries (about a specific offer, domain, or fan page), the data is complete. For broad, aggregate queries across the whole database — like “top 10 over the last six months” — the AI analyzes a sample of the highest-reach cards rather than 100% of the database, so the bottom of the ranking may shift slightly. For a precise breakdown, narrow it down by geo, vertical, or time period.
For example, we asked for the top offers over the past six months:
And got the top 10 nutra offers from March 28 to September 28. In this example, the ranking is built on creative reach — there’s no view or like data in the database, so reach is currently the only usable ranking metric:
Right after the answer come the analytics and notes, along with info on how many views were spent and how many remain on your balance.
In this example, the whole query used about 330 views — less than 1% of the 30,000-view monthly limit on the Basic plan:
The query itself took about 2 minutes, which is faster than manual analysis. Try MCP in your own MTWSPY account — connect Claude, ChatGPT, or Gemini using the instructions above, and your first aggregate query across the ad database will take you a couple of minutes instead of manually sorting through cards.
Want to try MCP in MTWSPY? Use promo code WELCOME for 25% off your first purchase on any plan.
FAQ
Does MTWSPY spend views?
Yes, views are deducted, since the AI “views” the ads.
Can I connect any AI, or only Claude?
You can connect any AI that supports MCP.
How do I make sure the AI only draws on the MTWSPY database?
Just mention in your request that you want the answer to come from MTWSPY. For example: “Analyze gambling creatives from MTWSPY over the last six months.”
Does querying the MTWSPY database use up AI tokens?
Yes, it uses the standard number of tokens, same as any regular request.
How many views does one MCP query use?
It depends on how broad the query is: a narrow question about a single offer or domain costs a few views, while a broad, aggregate query across the whole database costs noticeably more. For example, the top-10 nutra offers over six months in the example above used about 330 views.
Why is the offer ranking based on reach instead of views or likes?
The MTWSPY database doesn’t track likes or views for ads — it stores reach. So for aggregate queries, the AI ranks by reach, since that’s the most complete and reliable metric available.
What if my API key is lost or exposed to someone else?
Contact MTWSPY support on Telegram — they’ll revoke the key and issue you a new one.