Superwall Keynote 2026: Let Your Agent Run Your Monetization

This week we hosted Superwall's first all-day event at NY App Week, with our customers, our partners and a lot of people I've learned from over the years in one room. I gave the keynote, and you can watch the whole thing above. This post covers the same ground, with the slides.

The editor we built for humans

About a year ago we were in a sticky spot. We'd spent four years building the best paywall editor for humans, and in that same year the best engineers we know started handing their work to agents. So did we.

People kept using Superwall. They just had less fun doing it. For a long time Superwall solved two problems: it made paywall experiments easy to launch, and easy to measure. For the first half of this year, if I'm honest, we only made them easy to measure. Dragging and dropping in an editor isn't easier than chatting with Claude. And reading a dashboard isn't an easier way to understand your metrics than asking an agent to explain them.

So about six months ago we reoriented the whole company around one job: helping your agents make you more money.

We didn't come up with this on our own. It started with a tweet from Razeen, who reverse engineered our editor and wrote a skill so Claude could edit his paywalls. The next day, Duncan on our team opened a PR that put the same thing into our own editor: Superwall's first AI paywall editor.

That was more convenient than using the editor yourself, but it wasn't the same thing. An agent inside our editor still couldn't see what only lives in your codebase, like your branding and how your product works. So a month later we launched the Superwall MCP, which lets your agent connect to Superwall to build paywalls, launch experiments and analyze the results.

Then customers started doing things we hadn't planned for. Pritesh at Status AI wired the Superwall MCP into Paperclip, an open-source tool for running a team of agents, and automated their monetization from experiment idea to live paywall. A lead agent plans and delegates, an analyst reads the data and writes the next experiment, a paywall operator builds it in Status's design system, and an archivist keeps the history. Every morning at 9 the loop decides which tests keep running and which stop. Pritesh calls himself "the board": the system asks for his approval in Slack before anything ships.

That's when we realized we were only getting started.

Agents will run your monetization

Our core belief at Superwall is that within the next 12 months, agents will run your entire monetization and experimentation pipeline. More literally: the same coding agents you use to build your product will own the funnel from install to purchase.

So why isn't every customer already running their app in an automated optimization loop? It isn't that the agents aren't smart enough. This year Claude made more headway on the Riemann hypothesis in 36 hours than mathematicians had in 40 years.

The models are ready. Your tooling isn't.

In almost every industry, people are using AI to push the frontier. Building tools that point those agents at the rest of our work is a story that's still being written. We aren't writing the whole book, but we're proud to write the chapter on apps, because apps are what we know best.

Superwall is in more than 12,000 apps on the App Store, up 112% year over year. In the last 12 months, almost 200 million monthly active users saw more than 2.3 billion paywalls on Superwall, and they generated over $1.5 billion in revenue. We know which levers move monetization and how to design products around them, and we know the process well enough to help you automate it. You focus on the product; your agents make you more money.

The loop, and what an agent needs to run it

Monetization is a loop, not a launch. The rules are simple:

1. Run a paywall or onboarding experiment.
2. Split users into the ones who converted and the ones who didn't.
3. Find the patterns in each group.
4. Build experiences around those patterns.
5. Repeat.

An agent can run that loop every day, if it has three things:

• A heartbeat. It runs on its own, unprompted and on a schedule.
• Raw data. It can see what every user actually did.
• Full control. It can change the whole experience, from install to purchase.

For most apps, the onboarding and the first paywall a new user sees account for 80–90% of revenue, based on Superwall's onboarding funnel data. So that's where we started.

Full control starts with code: no more editor

The first question was how to give your agent complete control from install to paywall. The answer wasn't our editor. Giving agents tools to edit the JSON behind a page was neat, but JSON isn't code. So the next version of Superwall ships without a point-and-click editor.

• Paywalls live in your codebase. They're React and TypeScript, in git, next to the rest of your product, where your agent can see them. That's the Superwall Framework.
• Superwall hosts them. You push a paywall the way you deploy a website to Vercel.
• Skills come bundled. Superwall's skills teach your agent to build well-thought-out onboarding flows with rich animation. They feel better than native.
• They're websites, too. The same code that runs in your app runs your web2app and app2web flows.

We aren't deprecating the editor for humans. When you're ready, your agent can migrate your editor paywalls in a few minutes.

Because the paywall lives next to your product, your agent can use your own backend in onboarding. In Magic Edit, users upload a selfie and get a preview of what the app will do with it. Christo on our team built a whole 3D scene in three.js that shows which options you've picked. And Ian put Doom in his paywall, just to show off.

Raw data: the SQL API

Next, we wanted agents to do the same detective work a person does when a metric moves. Charts and breakdowns are great for trends, but they don't tell the whole story.

So we launched the SQL API. Every agent using the Superwall CLI or the MCP can query the same production database that powers our charts. That includes:

• Every event you send us, including all your custom events.
• Every event we track for you, like paywall opens and subscription lifecycle events.
• All your user and device attributes.

That means Claude can follow what every single person did, whether they converted or not, and you'd be surprised what turns up. In Magic Edit, we found that most people who cancelled their trial hit an error in their first session. Claude didn't need creativity to look for that. It read every event from every user who cancelled. When you have the raw data, you don't need a human's intuition to spot the pattern.

Raw data starts at the ad

The user's journey doesn't start at the install. It starts when they see your ad, and you can't optimize for new users if you don't know where they came from. So we built attribution into Superwall. Here's how it works:

1. Superwall gives you unique links to run web campaigns.
2. Someone taps your ad and lands on the App Store.
3. Superwall forwards the click details, and the ad platform makes the match.
4. We send every revenue event back to the ad platform.
5. We make the attribution data available inside your app.

We launched with Meta. TikTok, Google Ads, Snapchat and ChatGPT ads are coming in the next few weeks. With attribution in the app, your agent can tailor onboarding to the ad each user came from.

What the loop looks for

The pattern-finding step works best when you know which answers predict revenue. We ran a meta-analysis of more than 400 apps on Superwall to find the onboarding questions worth segmenting on. The spread is the gap in ARPU between the best and worst answer to each question.

Question	Spread
How old are you?	2.89×
What do you do for work?	2.64×
How did you hear about us?	2.03×
By when do you need it?	1.92×
Who is this for?	1.85×
Which of these apply to you?	1.78×
Where do you shop? / What do you spend today?	1.72×
Height, weight, target weight	1.69×
Gender	1.69×
What's your main goal?	1.64×
How often will you do this?	1.62×
Which topic or content?	1.60×
Relationship, kids, pregnancy, child age	1.57×
How big is the problem today?	1.54×
What have you tried before?	1.54×
How experienced are you?	1.42×

Say someone tapped a weight-loss ad, told you they want to lose weight, and still didn't convert. That's a problem. It's also an opportunity, if you have the tools to act on it.

How much can one fix like that matter? Say weight loss is 10% of your installs, and you lift that funnel by 50%. That's only 5% overall. But it might be what pushes that creative to a positive ROAS, and then you can scale it. Do that again and again, and small wins compound into big ones.

Full control, all the way to checkout

A side effect of putting paywalls in code is that the onboarding you optimize in your app can run anywhere.

Web2app. With the same code, your agent can push a flow to the web, so people can check out online, before or after they install. Superwall handles redemption in the app, and hosts the web restore flow and the subscription management portal.

App2web. Since app2web checkout was allowed in the US, a lot of our customers have used it to their advantage. In Superwall's early data, initial conversions dropped 11.9%, but net proceeds rose 19.85%. The best app2web flows let people pay on the web while it still feels like they never left the app.

Onboarding flows are just web pages, so Superwall can blur the line between where the app ends and the web begins. That's why we built Teleport: your agent can open the same in-app paywall or onboarding flow in Safari. As markets outside the US follow the US on regulation, checkout methods and paywalls keep fragmenting, and one workflow for all of them goes a long way.

Your agent runs the loop, and you approve it

Superwall's skills teach your agent how to do all of this. It's still a little early to let an agent run without a human in the loop, so whatever you do, review experiments before you green-light them. But lean on your agent and these skills as much as you can. If your agent isn't running your monetization, you're ngmi.

Or we'll run it for you

If you'd rather not run the loop yourself, we'll do it for you. Tangent is the newest part of the Superwall family (welcome, Vahe!). Their team of experts audits your account, runs an experiment every week, ships the winners and repeats. They work with every major setup. Half of their customers don't even use Superwall: they build paywalls in whatever tools you use, and they'll even open the PRs. Across more than 100 apps, they've added over $50M in revenue from more than 5,000 experiments, with a 30–40% lift in ARPU.

The Tangent team's ideas are behind nearly all the experimentation skills we'll release later this year. You can read more about working with them.

Everything else we shipped this year

• The Superwall iOS app
• The Web SDK
• The Unity SDK
• Webhooks v2 and the subscription status API
• An entirely new dashboard and design
• The in-app Customer Center

Pricing, spelled out

Our pricing is simple, and surprisingly free. You only pay for revenue Superwall generates. Revenue we only track is free. Unlimited MTR, integrations, charts, webhooks, cross-platform subscription status, revenue tracking and the in-app purchase SDK cost nothing. For example, an app making $50k a month, with half of it going through our paywalls, pays $250. The details are on our pricing page.

Switching has never been easier, either. We're launching an import tool that brings in your products, entitlements, purchase history and more from other providers in one click.

One more thing: Supervibes

I've been working on a side project for over a year. Interest in building apps is exploding, and we don't think it'll slow down. But in our own data, fewer and fewer of the people who sign up for Superwall actually ship. Based on Superwall's onboarding funnel, 70% of new users are working on their first app ever, and only 10% of those apps make it to the App Store. These are people with paid Apple developer accounts.

They need more help, so we built Supervibes: an all-in-one agent for building apps that actually make money. It comes with more than 20 skills, CLIs and MCPs tuned for building high-quality apps fast.

• A hosted simulator, streamed to your browser from a real Mac.
• Custom backends, databases and more, built and deployed without signing up for more accounts.
• An AI gateway, so your app can use any model for features like image recognition and agent workflows.
• Debugging tools that let Supervibes read logs on your device and in production, so it can fix almost anything.
• Install the app you built straight to your phone, with an Apple developer account.
• Submit to TestFlight and the App Store.

Thank you

Thanks to everyone who came, and to everyone who watched. A special thank you to the Superwall team and our partners: none of this happens without your work. The keynote is also on X if you'd like to share it.