MCP Deep Ad Analysis is a free, open-source agent skill that reviews your live Google, Meta and TikTok Ads data through the Flyweel MCP server. Ask how your ads are doing in plain words. The skill queries your connected accounts, compares each number with the prior period and tells you what should change.
It is one of six skills in the Flyweel skills repository, released under Apache 2.0. It uses the figures the ad platforms report, which makes it the fastest view of what is working.
Add the Flyweel MCP server to your agent and sign in to Flyweel when it asks. There is no key to copy. Follow the MCP setup guide or the page for your agent: Claude Code, Cursor, Codex CLI, Gemini CLI.
New to Flyweel? Create a free account first. No credit card required.
npx skills add flyweel-production/flyweel-skills --skill flyweel-mcp-deep-ad-analysis
Or install all six Flyweel skills:
npx skills add flyweel-production/flyweel-skills
In Claude Code, the plugin installs the skills and adds the MCP server in one go:
/plugin marketplace add flyweel-production/flyweel-skills
/plugin install flyweel@flyweel
Restart your agent after installing so it picks up the skill.
Connect at least one ad platform in Flyweel (Google Ads, Meta Ads or TikTok Ads), then try:
How are my ads performing?
Six playbooks cover the most common questions:
Results and spend by platform for last week against the week before, plus your top five active campaigns by spend. You get the change, the best and worst campaign, and one thing to do.
Spend and results by campaign over the last 30 days, with a 90-day weekly trend. It flags campaigns that take real spend and return no results, and campaigns whose cost per result rose three weeks in a row.
Spend, results and cost per result by platform. Platforms are compared only on the same result and in the same currency. If they differ, the skill shows them side by side and says why they cannot be ranked.
One campaign by day for 30 days, and by week for 90. It looks for a learning period after launch, a slow rise in cost per result and day-of-week swings.
Groups spend by the domains your ads send traffic to. A domain you do not recognise can be a sign the ad account has been taken over, so the skill raises it at the top of the answer.
Ad results in one query, deals created, deals won and revenue by source channel in a separate CRM query. Then ad spend per won deal and revenue per dollar of spend. A recent month with few wins is shown as still open, not failed, because deals take time to close.
| Tool | Use |
|---|
| get_setup_status | Called first. Shows connected platforms, selected accounts and sync state. |
| list_organizations | When you belong to more than one organisation. The skill asks which one. |
| list_ad_accounts | Account names, selection state and last sync time. |
| query_metrics | All analysis, with related queries grouped into one call. |
| get_connection_status, get_sync_status | Quick checks on a connection or a running sync. |
| connect_ad_platform, select_ad_accounts, trigger_sync | Setup changes, used only when you ask. |
- Named results over “conversions”. Platform conversions mix leads, installs and button clicks. The skill uses the named result that matches what your business sells, such as leads or registrations.
- Currencies stay apart. Spend in different currencies is never added together or converted to make a total.
- CRM gaps are a setup step, not zero deals. Without a connected CRM, the skill says so once, shares the connection link and answers from ads data.
- Sample size rules. Under 100 clicks, no conclusion from click rate. Under 30 results, cost per result is a rough guide. Under 10, no ranking. Under 7 days, no trend.
- Your history is the benchmark. Campaigns are judged against your own account, not industry averages.
Each answer leads with the most important finding and its number, followed by a table of your result, spend and cost per result against the prior period, broken down by platform. Then what changed, numbered next steps, and data notes: the last sync time, which result metric was used and why, anything not connected and any sample-size warnings.
- “How are my ads performing?” A review against the prior 30 days.
- “Give me a weekly check.” Last week against the week before.
- “Which campaigns are spending the most with the fewest results?” Budget efficiency audit.
- “Compare Google, Meta and TikTok cost per lead for the last 90 days.” Channel comparison.
- “Where are my ads sending traffic?” Destination check.
- “What did my ad spend bring back in closed deals?” Full funnel, with a CRM connected.
This skill reads what the ad platforms report. What your Flyweel Cards1 paid can differ by tax, fees and timing. For questions about cash, the card skills in the same repository take over:
- flyweel-real-cash-payback: what the cash your cards paid brought back.
- flyweel-where-to-spend-next: add, hold, trim or cut, per platform, with an amount.
- flyweel-card-top-up: how much to top up and when.
See all six in the Flyweel agent skills docs or read the launch announcement.
Want to analyse ad data without connecting accounts? Try the Ad Performance Analyser. It works with CSV exports, screenshots and pasted numbers, no account required. Learn more about MCP and how it works, or follow the MCP setup guide.
MCP Deep Ad Analysis FAQs
Common questions about live ad analysis with the Flyweel MCP server, setup, and connecting your ad platforms.
What is Flyweel MCP Deep Ad Analysis?
Flyweel MCP Deep Ad Analysis is a free, open-source agent skill that reviews live Google Ads, Meta Ads and TikTok Ads data through the Flyweel MCP server. It covers weekly and monthly reviews, platform and campaign comparisons, wasted spend, campaign deep dives, where your ads send traffic, and ad spend set against CRM deals. It uses the figures the ad platforms report.
What is a SKILL.md?
A SKILL.md is a markdown file of instructions an AI agent loads when your question matches it. It tells the agent which tools to call, which figures to trust and how to answer. The Flyweel skills are open source, so you can read exactly what your agent will do before you install them.
What is MCP (Model Context Protocol)?
MCP (Model Context Protocol) is an open standard that lets AI agents connect to outside data and tools. It lets Claude Code, Cursor, Codex and other agents query live data, such as your Google, Meta or TikTok campaigns, without manual exports. The Flyweel MCP server provides ad platform, CRM and card data through this protocol.
How do I get started?
Three steps: (1) Connect the Flyweel MCP server to your agent and sign in to Flyweel when it asks. (2) Install the skill with npx skills add flyweel-production/flyweel-skills --skill flyweel-mcp-deep-ad-analysis and restart your agent. (3) Connect at least one ad platform, then ask how your ads are performing.
Do I need a Flyweel account?
Yes, a free Flyweel account is required for the MCP server connection. Sign up at signup.flyweel.co/mcp, no credit card required. The account connects to your Google Ads, Meta Ads and TikTok Ads so the skill can read live campaign data.
Is it open source?
Yes. It ships in the flyweel-production/flyweel-skills repository on GitHub under the Apache 2.0 licence, alongside five other Flyweel skills.
How is this different from the Ad Performance Analyser?
MCP Deep Ad Analysis reads live data through the Flyweel MCP server, so you export nothing. The Ad Performance Analyser works from data you paste or upload (CSV exports, screenshots, API replies) and needs no account. Use MCP for live, connected analysis; use the Analyser for an export you already have.