Ad Superpowers connects your advertising, analytics and tracking data to the AI you already work in. A question spread across eight tabs today becomes one question. And you can act on the answer, once you approve it.
From question to evidence to execution. In one place.
The accounts are there. The AI is there. What is missing is the connection, so you do the joining by hand, every single time.
Every platform has its own interface and its own export button.
A question that touches two sources costs you half an afternoon.
What you worked out last month sits in a sheet nobody can find.
It only starts once you have pasted.
It does not know your client, only what you typed in this chat.
Changing anything in an account is out of the question.
Not another dashboard. A connection that opens your accounts to the AI you already use, so one question reaches every source at once.
And when that AI is allowed to change something, a person belongs in between. That is where the real work sits.
No extra dashboard. You connect your accounts once, then you talk to your own AI and it has full access to everything happening in those accounts.
Standard OAuth on eight platforms. No integration project, no middleman, no shared passwords.
In plain language. The system reads every platform at once and gives you one answer with context, instead of eight tabs.
If you want to. Every change is shown first, you approve it, and only then does it happen. All of it is logged.
The advertising side, the analytics side and the tracking side. Available in one go.
You are not tied to one vendor. Connect once, use it wherever you already work.
Claude Code has a plugin of its own: tools, skills, workflows and agents in a single install.
They are free and they do what they promise. But a connector links one platform to your AI. That is a bridge, and what drives across it is still yours to work out.
If every platform ships one tomorrow, you have eight that do not know about each other. The question of why your cost per click is rising while your rankings drop touches three sources at once.
A tool tells you what is possible. A skill tells you what to do. 120 skills with benchmarks, plus 29 workflows that write out a method.
A connector has no idea which client you mean, what the budget is, or what went wrong last month. You explain it again every conversation.
There, a change goes through immediately. Here you see what will happen first, you approve it, and who approved it is recorded.
Meta's own connector has 82 tools and cannot upload video. Ad Superpowers goes from a URL to a paused ad in one step.
Counted in the official schema export of 21 July 2026. 24 specialist tasks measured side by side.
This is not a roadmap. It works today, with the tools and workflows already in the product.
Keyword Planner gives volume, ideas and competition per market and per language. The base under any proposition.
Search Console gives the real queries, positions and click-through rates. Not a panel, actual demand.
Analytics 4 for on-site behaviour. TikTok and Meta for audience make-up and interests.
Creatives, messaging and auction insights from the channels you have access to.
A workflow is a research method written out. The AI follows the steps a specialist would take.
The conversation below is an example. The numbers in it are made up. Every other number in this deck is not.
You ask the way you would ask a colleague. The system translates it to the right sources.
Search data, behaviour and creative side by side, with a note of where each number came from.
The same place builds the campaign the research points to. Once you approve it.
The tools are the surface. The skills are where the expertise sits.
Eight integrations turn separate accounts into one searchable whole. 34 read, 17 write.
29 worked-out workflows for multi-step jobs. 120 skills with benchmarks and decision rules for the AI to draw on.
Creative analyst, media buyer, marketing strategist, reporting agent and tracking specialist.
Most AI tools stop at reading. On six of the eight platforms you can actually change something here, with an approval step in front of it.
Analytics 4 and Merchant Center have no write actions worth opening up. LinkedIn writes run in a closed beta: the app holds edit rights for a limited number of ad accounts until the expansion is approved.
A write action is not a button but a four-step conversation. This path runs on Meta today. At the end there is a record: which plan, confirmed by whom, when, and with what outcome.
The system does nothing yet. It works out what would happen.
Every field that will change is in it. So is what it leaves alone, such as the budget.
For example PAUSE-4F2A9C. Change the plan and it takes a different token. After thirty minutes it expires.
Ask for the same thing again and the existing record comes back, not a second ad. If it fails halfway, the record says what did and did not go through.
At the end of the month you do not have to reconstruct anything from chat history. You have a list.
Per workspace and not per person, because accountability only the author can see is not accountability. The activity log with export runs today; the full record per individual action is still being built.
No more account numbers, but clients with a budget, a goal and a history. If you run one brand instead of twenty clients, a client profile is simply your brand profile.
Budget, currency and the overall goal.
Targets per channel, with metric and threshold.
Which ad accounts belong to which client.
Watch-outs and learnings, in plain language.
The account manager responsible.
Every change is versioned and reversible. Deleting is only possible from the dashboard, never from the AI.
Someone writes down once how you look at a brand. That way of working becomes a skill, and you roll that skill out across the team.
Write or upload them in the dashboard, or straight from the conversation with your AI.
Keep them personal or publish them across the workspace. Two separate levels, chosen per skill.
Duplicate, adjust, withdraw or archive them when they no longer hold.
Your own brand model becomes the model the system uses.
It stays yours. You can export all of it in bulk and take it with you.
You do not want twenty clients in one big bucket, and you do not want twenty separate accounts with twenty invoices either. A workspace is the middle ground: walled off per client, under one subscription and one login.
A walled-off environment per client or brand, with its own team, its own connected accounts and its own permissions. What happens in one workspace is not visible in another.
One person connects an account, then an admin assigns it. A member sees only what is assigned to them, an admin sees everything. Who granted the assignment is recorded, and write permission is set per person.
Every member creates their own API key, separately per workspace. An admin sees every key in the workspace, a member only their own. Revoking one leaves the rest untouched.
Every write action lands on the person behind the key that was used, in the log of the workspace where it happened. Filterable by period and by person, and exportable.
Your colleagues do not have to connect anything themselves. You connect, you assign, and you can see who did what.
If someone moves off the client, you lose one workspace and not your whole setup. How many workspaces you get depends on your plan.
The EU AI Act applies in full from 2 August 2026. We were not built around it as an afterthought, we were designed around it. How we use AI is described publicly at adsuperpowers.ai/ai-transparency.
Every write action shows exactly what will change. It only happens once you approve it. Never quietly.
Standard connections from Meta, Google, LinkedIn and TikTok. Each Google product is connected separately, so you never grant more access than you mean to. The owner can revoke it at any time.
Data goes back to the AI of your choice. Nothing is trained on it and nothing is stored long term.
Every call logged, tied to the person who approved it, in the log of the workspace where it happened. Exportable, so it holds up towards your client and towards a regulator.
Not a wish list. Every track is a set of tickets in our planning, and the label next to it is the real status rather than the convenient one. What is already finished is not on this list.
We measure our own workflows against a catalogue of specialist tasks per platform. Meta has been measured and the results are in the product. The other platforms follow. Not claims, measurements.
The per-workspace log runs today, with export. What is still open is the full record per individual action and the monthly statement drawn from it.
Hard limits per client: a maximum budget change in percent, an absolute ceiling, and a cap on the number of changes per account per day.
One searchable library across Meta, TikTok and LinkedIn. See what a competitor is running without digging platform by platform. Waiting on the TikTok API to come back.
Amazon Ads and ChatGPT Ads as the next connections. Each has its own approval process at the platform itself, so the order is not entirely ours to set.
This is not a tool that makes your work faster. It is the layer that lets you take on the things you leave undone today, and a place to try what you would otherwise never look into.
A competitive analysis across four channels. A keyword gap across two languages. Checking per client which creative is worn out. You can do all of it, it just costs a day, so you do not.
If an answer takes five minutes instead of an afternoon, you ask ten instead of one. You follow what catches your eye, instead of deciding up front what is worth the effort.
Work out what a shift would do before you make it. Test an assumption without putting budget behind it. Being wrong costs you a conversation instead of a week.
The fine-grained work still gets done, just not by you. What is left are the questions that need judgement, and that is what you were hired for.
The point is not that you work faster. You start doing different things.
Every month brings more tools that can do something with advertising. Almost none of them give you the control you need to run client work through them. That is what we are building, from Europe.
AI that moves budgets on its own is not a product, it is a liability. Every proposal passes a person who sees it and approves it, with a trail of who did what and when.
Client budgets, targets, learnings, your own way of working and who may reach what. In one workspace, instead of scattered across heads, folders and separate chats.
Built to European standards for oversight and accountability, with the EU AI Act as the starting point rather than an afterthought. Ready for teams anywhere.
Take a question you normally outsource, or one that normally costs you an afternoon. If the answer beats what you do today, you know enough. And if it does not, you will know that within a week.
What would you ask if your AI could reach everything?
For people working solo. 1,500 tool calls, 25 connected accounts, 1 user, 2 workspaces.
For agencies. 6,000 tool calls, 100 connected accounts, 5 users, 5 workspaces, roles per person.
Unlimited workspaces, custom integrations, an SLA and guided onboarding.
Paying annually brings that down to €79 and €199 per month. Code FOUNDING30 takes another 30% off for as long as you stay. Prices exclude VAT, cancel any time. Failed tool calls do not count.