Video Summary

WebMCP: Let AI Agents pay you money

Greg Isenberg

Main takeaways
01

WebMCP exposes concise, agent-friendly tools in the browser so agents can read and act on pages without scraping the DOM.

02

The browser session carries login state, keeping tools conditional and avoiding API key/token handoffs.

03

WebMCP sits between headless APIs and in-app agents: it preserves the visual UI while letting the user's agent bring context and memory.

04

Early high-value markets: compatibility-driven commerce, SaaS admin consoles, regulated read-only flows, and internal tools.

05

Two practical startups: a WebMCP conversion agency (setup + monthly monitoring) and an agent mystery shopper subscription service.

Key moments
Questions answered

What is WebMCP?

WebMCP is an experimental browser/UI feature (from Google and Microsoft) that exposes a short list of agent-friendly tools on a website so AI agents can read the page and perform actions without scraping the full DOM.

How does WebMCP handle authentication and permissions?

WebMCP relies on the browser session for auth: the user's logged-in session is shared with the agent, so tools can be conditional on login state and API keys or tokens don't need to be passed to the agent.

Where does WebMCP sit relative to headless APIs and in-app agents?

WebMCP sits in the middle: it preserves the visual web UI users prefer while exposing programmatic tools so external personal agents can act without running headless APIs or being embedded as vendor in-app agents.

Which markets are expected to benefit first from WebMCP?

Initially compatibility-driven commerce (complex stores and accessories), SaaS admin and analytics consoles, regulated read-only flows (insurance, banking), and internal tools for enterprises and startups.

What two startup ideas were suggested to monetize WebMCP now?

1) A WebMCP conversion agency that makes business sites agent-ready (one-time setup fees ~$2k–$10k plus monthly monitoring). 2) An agent mystery shopper subscription that tests whether agents can complete key journeys and reports conversion risks (recurring fees ~ $100s/month).

What did the espresso-store demo demonstrate?

The demo showed an agent comparing two machines, checking counter width and accessory compatibility, adding an item to the cart, and applying a coupon — all via WebMCP tools rather than DOM scraping.

The Rise of AI Agents and WebMCP 00:00

"AI agents are about to become a new kind of visitor on the internet, and I believe billions of dollars of wealth will change hands in the process."

  • AI agents are emerging as a significant presence on the internet, leading to transformative changes in online commerce and interaction.

  • It’s anticipated that millions of AI agents will engage in browsing and purchasing, which presents a challenge since current websites are not optimized for AI interaction.

  • There is a need for websites and apps to be designed specifically for AI agents, raising the question of what features these sites should have to facilitate such interactions.

Introducing WebMCP 00:28

"WebMCP is a new experimental feature from Google that allows websites to expose cleaner actions for AI agents."

  • WebMCP simplifies AI interaction by allowing websites to clearly define what actions an AI agent can take, rather than having agents decipher human-oriented interfaces.

  • This new feature is poised to revolutionize how AI agents operate on websites, providing them with actionable instructions like how to search, book, or purchase directly.

  • The host plans to discuss the distinctions between WebMCP and traditional methods, why this innovation is significant, and explore potential startup opportunities that leverage this technology.

Practical Applications of WebMCP 02:30

"WebMCP allows for agent-native websites, which can significantly reduce the need for inefficient browsing and scanning."

  • A personal example illustrates the practical use of WebMCP in a shop specializing in espresso machines, showcasing how it can streamline decision-making for consumers.

  • Rather than manually sifting through information, an AI agent could directly interact with a WebMCP-enabled site to compare products based on specific criteria like dimensions and functions.

  • WebMCP enhances the efficiency of interactions, allowing agents to provide tailored recommendations and streamline the purchasing process.

The Efficiency of WebMCP Over Traditional Methods 07:24

"WebMCP enables websites to provide action buttons for agents without them needing to parse the entire DOM."

  • One of the significant advancements with WebMCP is that it eliminates the cumbersome process of scraping and parsing website code, enabling a more direct and user-friendly interaction model.

  • Websites can design a set of tools that are easily accessible and conditional based on user login status, allowing for a more tailored experience without complex credential management.

  • This innovation creates a flexible approach for building agent-native applications, making it easier for businesses to adopt new technologies and optimize their web presence.

Understanding Headless Applications and Agents 09:42

"Headless means running an app via your agent without ever seeing the app; there's no app in the browser, no website."

  • Headless applications operate without a traditional user interface, allowing agents to interact with APIs directly. This presents a unique method of application use, focusing on backend communication without visual components.

  • Users can implement their own agents to interact with a headless API, but they must manage authentication, such as API keys, which adds a layer of complexity to the setup.

  • This approach can complicate development but provides flexibility in how applications and agents communicate.

Exploring Different Agent Interfaces 10:42

"The in-app agent is the vendor's agent that interfaces with the app's logic and data."

  • An in-app agent requires the user to configure everything, including authentication tokens, and is particularly tailored to operate within the confines of a specific application.

  • WebMCP serves as a middle ground that alleviates users from the burden of token management while allowing them to utilize their pre-existing agents, thus streamlining the user experience.

  • As consumers increasingly adopt personal agents, WebMCP helps maintain a seamless integration with existing tools without necessitating major structural changes to the applications.

Practical Examples of Agent Interactions 16:06

"We're going to want them to go and do stuff for us, ensuring that they can complete their tasks seamlessly."

  • Personalized agents can assist consumers with complex purchasing decisions, helping to navigate intricate product specifications and facilitate the purchase process.

  • In an example related to selecting coffee equipment, an agent is capable of contextualizing user needs and suggesting products based on prior orders, enhancing the shopping experience.

  • Users can trust agents to efficiently add items to carts and apply discounts automatically, eliminating tedious manual tasks during online shopping, which demonstrates the beneficial integration of WebMCP in everyday scenarios.

Compatibility-Driven Commerce and WebMCP Tools 19:12

"You could do it in SAS admin, so any kind of admin board, console, or analytics tools can make it easier for your marketing team to deal with these things."

  • Compatibility-driven commerce can enhance various systems including camera setups and complex web stores. Implementing WebMCP tools can streamline processes in settings like SaaS admin boards or analytical dashboards, which benefits marketing teams and improves feature setup.

  • In regulated industries such as insurance and banking, using WebMCP tools can allow for a 'read-only' setup, providing users access to essential information without compromising sensitive data.

  • Internal tools can also be enriched with WebMCP solutions, catering to internal operations for entrepreneurs. This in-house use presents significant potential by creating operational efficiencies for employees.

Business Idea: WebMCP Conversion Agency 20:54

"The big idea is how can you make boring business websites agent-ready?"

  • One viable business concept is establishing a WebMCP conversion agency that focuses on transforming traditional websites into agent-ready platforms. This addresses the needs of boring businesses, such as law firms and HVAC services, by enabling functionalities like quote requests and consultation bookings.

  • The agency could offer a setup service priced between $200 to $10,000, with ongoing monthly fees for maintenance and monitoring. This model would create a steady cash flow by building retaining relationships with clients as they transition to working with AI agents.

Business Idea: Agent Mystery Shopper 24:26

"It's testing whether agents can complete the most important user journeys."

  • The second business idea is to create an 'agent mystery shopper' service that conducts tests on AI capabilities to complete key user interactions on business platforms, such as booking consultations or filing claims.

  • This service would generate reports on agent performance, highlighting issues like poor navigation or inadequate descriptions that hinder user experience. These insights could be turned into software solutions for continuous improvement.

  • This idea has the potential to charge clients a monthly fee for ongoing analyses and adjustments, merging well with the previously mentioned conversion agency business model.

Experimentation and Early Adoption 27:32

"There's a lot of potential here because we talked about this package... so that when it does get adopted, you're the first mover."

  • Embracing experimental technologies, like WebMCP, can provide a competitive edge when they become mainstream. Companies should explore and implement these emerging tools before they saturate the market, benefiting from early adoption.

  • The conversation emphasizes the importance of testing and iterating on experimental features, especially from major companies like Google, suggesting that early engagement with such tools can lead to substantial opportunities in the future.

Proposal and Launch Timeline 28:41

"They made the proposal two years ago and launched it at the beginning of this year."

  • The initiative being discussed has had a timeline that stretches back over two years. Initially proposed two years prior, it finally came to fruition with its launch earlier this year.

  • This timeline illustrates the planning and effort taken before actual implementation, highlighting how long it can take for ideas to materialize into active projects.

Personal Sentiment on the Initiative 28:47

"I wouldn't fade this personally, and I'm excited about it."

  • The speaker expresses a positive outlook regarding the initiative, indicating personal excitement and confidence in the project's potential.

  • Such enthusiasm reflects a belief in the value of the project, suggesting that it is worth monitoring and engaging with moving forward.