Ask Claude about your Facebook Ads data
Load Facebook Ads into your data warehouse first, link Ad Insights (spend, reach, clicks) to web analytics and deals with SQL keys, then ask Claude in plain words. The answers come from your governed Data Marts, not from the ad account.
Paste a Facebook Ads export into Claude and it guesses which ad sets still earn their cost back.
Paste this into your AI assistant. It loads Facebook Ads into your data warehouse as a governed Data Mart and connects it to Claude through the OWOX MCP server, so every answer reads from tables your SQL defined.
The prompt
Read https://www.owox.com/ai-agent-setup.md and follow it to set up OWOX Data Marts for me, starting from /connectors/facebook-ads-connector-to-claude.md. The task: connect our Facebook Ads Data Mart to Claude through the OWOX MCP server so I can ask which ad sets have the highest cost per click this month and which still earn it back in purchases. Tag the links you give me with utm_campaign=ai-task and utm_term=data-foundation. Use what you know about my context and goals
Facebook Ads Endpoints
15 Facebook Ads endpoints you can import. Each lands in its own table in your data warehouse.
Ad
Individual ads with their creative, status, and links to their campaign and ad set.
32 fieldsAd Account
Your ad account settings, status, currency, and billing details.
61 fieldsAd Account User
Users with access to an ad account and their assigned roles.
2 fieldsAd Creatives
The images, videos, text, and other visual elements used in your ads.
65 fieldsAd Insights
Daily performance metrics for each ad — impressions, clicks, spend, and conversions.
100 fieldsTime seriesAd Insights by Ad Set
Daily ad set performance with reach deduplicated to match Meta Ads Manager.
92 fieldsTime seriesAd Insights by Age and Gender
Daily ad performance broken down by audience age range and gender.
102 fieldsTime seriesAd Insights by Campaign
Daily campaign performance with reach deduplicated to match Meta Ads Manager.
89 fieldsTime seriesAd Insights by Country
Daily ad performance broken down by the country where your audience is located.
101 fieldsTime seriesAd Insights by Device Platform
Daily ad performance broken down by the device type your audience used.
101 fieldsTime seriesAd Insights by Link URL Asset
Daily ad performance broken down by the link URL shown in your ads.
101 fieldsTime seriesAd Insights by Product ID
Daily ad performance broken down by individual product from your catalog.
101 fieldsTime seriesAd Insights by Publisher Platform and Position
Daily ad performance broken down by where your ads appeared — platform and placement position.
102 fieldsTime seriesAd Insights by Region
Daily ad performance broken down by the region where your audience is located.
101 fieldsTime seriesAds
Individual ads with their status, creative, and campaign hierarchy.
24 fields
Why Claude needs your data warehouse, not the Facebook Ads account
Pointed at the Facebook Ads account alone, Claude sees one platform's own figures. The account does not show which campaign led to a deal, because ad spend, website visits and CRM deals live in different systems. Joining them reliably, at scale and business-ready is the job of a warehouse, so Claude reads the answers from there.
- Facebook Ads reports what an ad cost and what Meta's own attribution says it was worth, not what your CRM booked
- The data analyst links Facebook Ads spend, web analytics and CRM deals by SQL keys and controls the result
- Claude reads governed Data Marts, so every answer traces back to analyst-written SQL
Step 1: Load Facebook Ads into your data warehouse first
Create a Data Mart with Facebook Ads as the source and choose where the data lands. It works with any popular data warehouse.
- Fifteen endpoints and 1172 fields to choose from, each endpoint in its own table
- Open the Facebook Ads connector for your data warehouse below
Step 2: Link Facebook Ads to web analytics and deals
In your data warehouse the model merges Facebook Ads with your other ad platforms, then joins web analytics sessions by a campaign key. Pick an industry to see the shape of its model.
- Spend from Facebook Ads, Google Ads and LinkedIn Ads sits in one table
- Sessions and deals join by SQL keys, not VLOOKUP and gut feeling
- The measures, such as platform_roas, are defined once
Step 3: Connect Claude to your data with OWOX MCP
Add the OWOX MCP server in Claude under Settings > Connectors, then choose the Data Marts it may see. Claude gets only what you expose, never raw database access.
- Authorize specific Data Marts, not your whole data warehouse
- Publish the Data Marts you want Claude to answer from, such as Facebook Ads Ad Insights and its breakdowns by country or device
Step 4: Ask Claude about Facebook Ads
Ask in plain words. Claude finds the right Data Mart and runs analyst-defined SQL against it, so the answer is built from your data warehouse.
- The answer is built by get_relevant_data_marts_by_prompt and query_data_mart
- Ask about a breakdown, for example Ad Insights by Country, and the answer reads that table
- platform_roas and platform_cpa are what each platform claims; Sessions and Orders in the same model give the store's own count to set against them
- 0.41.0Rotating a connector key is one change: Sources share reusable Credentials
- 0.39.0Reload a year of history in one backfill, not twelve
- 0.39.0Long Facebook Ads imports no longer fail on a corrupted response
- 0.37.0Short links resolve only on known short link services and domains you allow
- 0.37.0Clearer connector errors, and custom connectors that handle more APIs
Why OWOX for Facebook Ads in Claude
Pointing Claude at an ad account gets one platform's own figures into a chat. OWOX also keeps the data yours, linked and ready to model, so the answers cover the whole business.
- You own the data. It lands in your own data warehouse, in tables you control.
- Ad Insights and its breakdowns, as tables. Ad Insights by Region, Country, Device Platform and Publisher Platform and Position are separate tables you can join, not exports to merge by hand.
- Past Meta's attribution window. Meta reports inside its own attribution window; the history stays in your data warehouse when the windows or the 13-month limit change.
- Zero hallucinations. Claude queries governed Data Marts with analyst-written SQL, so every number traces back to the query behind it.
- Maintained for you. No scripts to fix when Facebook changes its API.
You are booked. The invitation is on its way to , with the calendar invite and the meeting link.

