https://mcp.facebook.com/ads. It is remote-hosted by Meta and works with any MCP-compatible client, so there is nothing to install locally. You will be asked to authorise access to the ad account you want it to read.https://mcp.higgsfield.ai/mcp. This is the piece that actually generates the video and image output, and the quality of what comes out of it is the whole point of the workflow.The point of this step is to stop choosing creative by taste. Almost everyone running ads has a favourite that the data does not agree with, and the connector settles the argument in about ten seconds.
Look at my ad account and tell me which creatives are actually carrying performance over the last 90 days, not which ones I have spent the most on.
Rank them by the metric that matters for my objective, and for each of the top three tell me: what the creative actually is, how long it has been running, what its cost per result is against the account average, and whether it is still improving or has started to fatigue.
Then tell me plainly which single creative you would build variations from, and why that one rather than the others. If two are close, say so instead of picking arbitrarily.This is the competitor half, and it does not run through your ad connector. The Meta Ads Library is a public transparency tool. Anyone can browse it, no account and no login, and it shows you every ad a page is currently running.
Here is the part worth understanding properly, because it changes how you read it. The library does not show spend or performance for ordinary commercial ads. Only political and social-issue advertising carries public spend and impression data. So you cannot look up a competitor's best ad directly.
What the library does show you is how long an ad has been running, and that is the proxy. Advertisers switch off what is not working, quickly, because it costs them money every day it stays live. An ad that has been running for eight months is an ad that is paying for itself. That inference is the whole technique, and it is reliable enough to build from.
I am researching what already works in my niche before I make new creative. Here are the advertisers I want to look at: [paste 3 to 5 brand names or their Ads Library page links].
For each one, go through what they are currently running and identify the ads that have been live the longest, because longevity is the closest public signal we have to performance. Meta does not publish spend for commercial ads, so do not guess at budget or results, just tell me what has persisted.
Then break down the pattern across all of them: what the first three seconds do, what the offer is, how the ad is structured, and what they have in common that I should borrow. Tell me what is consistent across several advertisers rather than what one of them happens to do.Now the creative studio does the work. You are not asking it to invent something, you are asking it to rebuild a structure you already know performs, which is a far easier job to get right.
Take this winning ad structure: [paste the breakdown from the previous step, or describe the creative you want to build from].
Generate a set of variations that keep what is actually doing the work, which is the hook timing, the pacing and the offer, while changing the surface: the setting, the talent, the product angle and the visual treatment. Keep every output at commercial quality, because a rough variation tells me nothing about whether the idea travels.
Then give me each variation with a one-line note on what specifically I am testing with it, so I am running a real test rather than just making more videos. Finally, translate the strongest two into [list your target languages] keeping the same pacing, so I can run the same winner in other markets.One practical note on cost. Generating a batch of commercial-quality variations consumes credits, so decide how many you actually need before you start rather than generating fifty and picking two. The point of finding the winner first is that you need fewer swings, not more.