How to Earn 75% Revenue Publishing AI Agents: The Creator Economics Guide

Building an AI agent has never been easier, but getting paid for one is a different problem. Most of the advice on making money building AI agents focuses on the build itself: the prompts, the tools, the workflow. Far less of it covers the economics that decide what a creator actually keeps, such as where an agent is published, how revenue is split, and which costs are taken out before the split happens. For creators who recently lost a publishing channel, those questions are no longer theoretical. This guide walks through how agent marketplace revenue shares work, how to calculate your real earnings per run, and what to look for before you decide where to publish next.

Why Where You Publish Matters as Much as What You Build

An agent is only as valuable as the channel that puts it in front of users. A strong agent on a platform that stops distributing it earns nothing, while an average agent in a busy marketplace can earn steadily. That makes the choice of platform a business decision, not a technical one, and it deserves the same attention creators usually give to prompts and integrations.

Platform Risk: When a Publishing Channel Disappears

Agent.ai launched at INBOUND 2024 as what CX Today described as a professional network for AI agents, where builders could create agents and share them with a wide audience. Today the agent.ai address leads to BuilderPack, a HubSpot Labs toolkit that adds actions to HubSpot's Agent Builder. For creators who published there, the open community channel they relied on no longer works the way it did.

The Shift From Open Agent Networks to Closed Marketplaces

HubSpot's new direction is built around its own accounts. HubSpot launched Agent Hub and Agent Builder in public beta in July 2026 as one place to build and manage agents inside a HubSpot account. According to the HubSpot Knowledge Base, the Agent Marketplace only lists HubSpot's pre-built agents, and custom agents created in Agent Builder are not eligible for listing.

What Most Builders Never Get to Monetize

Even when an open channel exists, only a fraction of builders reach the point of publishing. Groath reports that agent.ai had around 2 million users as of INBOUND 2025, with about 26,000 building their own agents and roughly 1,800 agents published for community use. That is about 7% of builders, and it shows how much agent work never reaches a paying audience at all.

How Revenue Shares Work on Software Marketplaces

Revenue shares are not new. App stores and software marketplaces have used them for years, and their terms give creators a useful benchmark for judging any AI agent marketplace. What changes with AI agents is the cost structure underneath the split, which is why the headline percentage never tells the whole story on its own.

App Store and Shopify Benchmarks

The most familiar benchmarks come from mobile and ecommerce. Apple charges a standard 30% commission, reduced to 15% for developers earning up to $1 million a year through its Small Business Program. Shopify takes 0% on the first $1 million in lifetime app revenue and 15% after that, plus a processing fee. In other words, most developers keep between 70% and 85% of gross revenue on these platforms.

Gross Share vs Net Share: Why the Fine Print Matters

A revenue share can be calculated on gross revenue or on what remains after costs, and the difference is significant. A 75% share of gross revenue and a 75% share of the amount left after running costs are two different numbers. Before comparing platforms, check exactly what the percentage is applied to, which costs come out first, and whether any processing or listing fees sit on top of the split.

Why AI Agents Have Costs Traditional Apps Don't

A mobile app costs almost nothing each time someone opens it, but an AI agent spends real compute on every run. Anthropic reports that agents typically use about 4x more tokens than chat interactions, and multi-agent systems about 15x more. Those model and tool costs have to be covered somewhere, which is why agent marketplaces usually take execution costs out of each run before splitting what remains.

The Creator Economics Formula for AI Agents

Once you know how a platform treats costs, your earnings per run come down to a simple formula. It works for any marketplace that charges users per run, and it makes it easy to compare platforms that describe their splits in different ways. The inputs are the usage fee you set, the execution cost of the run, and the platform's percentage. Per-run pricing is also familiar to buyers now: Orb cites research showing 85% of surveyed SaaS companies had adopted usage-based pricing by January 2025.

Usage Fee, Action Costs, and Platform Fee

Start with the usage fee, which is what a user pays to run your agent. Subtract the action or execution costs of that run, such as model calls and tool usage. The platform then takes its percentage from what remains, and the rest is yours. Written as a formula: creator earnings equal the usage fee minus action costs, multiplied by your share of the remainder.

A Worked Example: What a 10-Credit Run Pays

Take a run with a 10-credit usage fee and 2 credits of action costs. That leaves 8 credits. With a 25% platform fee on the remainder, the platform keeps 2 credits and the creator keeps 6. The creator's 75% share applies to the 8 credits left after costs, which works out to 60% of the full usage fee in this example.

Pricing Your Agent So Margin Survives Scale

Because action costs come out first, a usage fee set too close to your execution cost leaves very little to split. Price with a buffer above your typical run cost, and watch for runs that trigger more steps than usual. Buyers are sensitive to pricing surprises: G2 found that 33% of recent AI agent builder reviews mention pricing as a dislike, so a clear, stable per-run fee also helps adoption.

What Former Agent.ai Creators Should Look For in a New Home

If you are moving agents from one platform to another, the checklist is short but strict. The goal is to avoid rebuilding your catalog on a channel that could close or restrict publishing again. These three criteria separate a durable creator platform from an internal agent builder with a marketplace label attached.

Open Publishing, Not Internal-Only Builders

Many agent builders are designed for a company to use its own agents internally, not for creators to publish to outside users. Before moving, confirm that custom agents can be listed publicly and that anyone on the platform can find and run them. If the marketplace only lists the vendor's own agents, it is a product catalog, not a creator channel.

Transparent Per-Run Settlement

You should be able to see what each run earned and why. Look for a platform that shows the usage fee, the execution cost, and the platform fee for every run, rather than a single monthly payout number. Clear settlement makes pricing decisions easier and lets you spot runs that cost more than expected before they eat into your margin.

Built-In Discovery and Buyers Already Browsing

A marketplace is only worth joining if buyers are already there. Demand is growing fast: MarketsandMarkets projects the AI agents market to grow from $7.84 billion in 2025 to $52.62 billion by 2030, with ready-to-deploy agents expected to hold the largest share. Choose a platform where your agent appears in the same place those buyers already browse.

How Caywork's Creator Program Works

Caywork was built as an open marketplace where creators publish agents for anyone on the platform to run, and earn each time they do. Its creator program covers the three criteria above: public listing, per-run settlement, and discovery inside a shared marketplace. Here is how the economics work in practice.

Build and Publish Without Upfront Costs

According to Caywork Creator, creators build agents in a visual interface with triggers, agent cores, pre-built tools, and logic nodes, with no coding required. The platform provides access to more than 20 AI models and over 70 integrated tools without requiring your own API keys. Building, testing, and listing an agent on Caywork is free.

The 75/25 Split, Explained Run by Run

Creators on Caywork set their own usage fee. For each run, action node costs are covered first, Caywork takes a 25% platform fee on the remaining usage fee, and the creator keeps the rest. The homepage example shows a 10-credit run settling as 2 credits of action costs, 2 credits of platform fee, and 6 credits for the creator. A dashboard shows spending, node costs, and your share on every run.

Move Your Agent to Caywork

If your agents lost their publishing channel, rebuilding them on a platform designed for creators is the fastest way back to earning. Start with your best-performing agent, set a usage fee with a buffer above its action costs, and track its per-run settlement for the first few weeks before moving the rest of your catalog.

---

Publish your first agent on [Caywork] for free, set your own usage fee, and keep 75% of every run after action costs.

Frequently Asked Questions About Making Money Building AI Agents on Caywork

The questions below cover what creators ask most often when deciding where to publish and how Caywork's revenue share works. They focus on the practical details that affect earnings, from fees to pricing. For the latest terms, the Caywork Creator page is always the best reference.

Does It Cost Anything to Publish an Agent on Caywork?

No. Building, testing, and listing an agent on Caywork is free, and there are no upfront fees. You also do not need your own API keys, because the platform provides model access and integrated tools. Costs only come into play when a user runs your agent, and they are covered from the usage fee.

Is the 75% Share Calculated on the Full Usage Fee?

No. Caywork covers action node costs first and then takes a 25% platform fee on what remains, so your 75% applies to the usage fee after action costs. In the 10-credit example, that means 6 of the 10 credits go to the creator. The lower your action costs relative to your fee, the closer your share gets to 75% of the total.

Who Sets the Price of My Agent?

You do. Each creator sets the usage fee for their own agent. A good starting point is to measure the typical action cost of a run, then set a fee with enough buffer that your share stays meaningful even when a run takes more steps than usual.

Can I Move Agents I Built on Another Platform?

Agents are not transferred automatically between platforms, but you can rebuild them on Caywork using its visual builder, pre-built tools, and supported models. Many creators start with their most-used agent, confirm it performs well, and then move the rest of their catalog once the pricing and settlement look right.

How Do I Know What Each Run Earned?

Caywork settles every run individually and shows the breakdown in a creator dashboard, including end-user spending, node execution costs, and your profit share. Total executions and usage logs are also available, so you can see which agents earn the most and adjust pricing based on real data.

Final Thoughts

Making money building AI agents depends as much on where you publish as on what you build, and a revenue share only means something once you know which costs come out first. Caywork gives creators an open marketplace, free publishing, a usage fee they set themselves, and a 75% share of every run after action costs, with each run settled in full view. If your agents need a new home, Caywork is ready for them.

References

• Anthropic, "How We Built Our Multi-Agent Research System"

• Apple, "App Store Small Business Program"

• BuilderPack, "AI Actions for HubSpot Agent Builder"

• Caywork, "AI Agent Marketplace"

• Caywork Creator, "Build an Agent. Get Paid When It Runs."

• CX Today, "HubSpot Introduces Agent.AI: The Only Network of AI Agents"

• G2, "What Buyers Really Think About AI Agent Builders"

• Groath, "The AI Agent Marketplace in 2026: When to Buy, When to Publish, and When to Ignore It"

• HubSpot, "Meet Agent Hub and Agent Builder"

• HubSpot Knowledge Base, "Install Agents From the Agent Marketplace"

• MarketsandMarkets, "AI Agents Market Worth $52.62 Billion by 2030"

• Orb, "40 SaaS Pricing Statistics"

• Shopify, "App Store Revenue Share"