Y Combinator’s slogan, “Make something people want,” never mentioned a product. It was about distribution. Build something that finds product market fit, and people come. For 20 years, that distribution was through search engines, social algorithms, paid ads, and word of mouth. All channels are directed by people.
That changed. In the past six months, Cloudflare, Shopify, Stripe, Supabase, Netlify, and Google have each invested in becoming agent-friendly, creating a new distribution channel: AI agents visit websites, extract information, compare options, and complete transactions on behalf of the people they refer. No one linked this. Six companies in different industries saw the same thing and built it independently.
When six companies in different industries create one category of visitors independently, the channel is real.
Your website still needs to do something that people want. But the agents now how many of those people get it. A website that works for people and fails for agents is a product with a broken distribution channel.
6 Companies in 6 Industries Made the Same Agent Infrastructure Bet
Cloudflare has dedicated a week of rolling out to agents in April 2026. Not a single feature announcement. A week. Agent authentication with Web Bot Auth with GoDaddy. Agent-readable content with Agent Markdown. Agent-driven tasks with WebMCP in Browser Runtime. Agent rating by Agent Readiness Score on isitagentready.com. An infrastructure vendor doesn’t clear its calendar for something speculative.
Shopify has deployed an Agent Toolkit so that any AI agent can browse the store’s catalog, check inventory, and complete checkout with a structured API. The seller does not build anything new. Google went further with the Universal Commerce Protocol, expanding it at I/O 2026 with the Universal Cart, and added the Agent Payments Protocol to the 60-member FIDO Alliance. The agent’s commercial layer went from draft to production integration in less than three months.
Projects are submitted through Stripe, a platform where AI agents can create accounts, purchase domains, use infrastructure, and manage subscriptions through Stripe’s payment channel. The infrastructure procurement layer became agent-transacable overnight.
Netlify built netlify.ai, a dedicated space where AI agents can deploy websites, manage projects, and access a platform full of agent capabilities. It is not a feature within an existing product. A separate entrance for non-human visitors. I spoke with their CEO about this last week, and the thinking was straightforward: If agents are going to use websites on your property, give them a front door designed for them.
Then there is Supabase. Their tagline is “Postgres development platform.” That says almost nothing to the vibe codes that have made Supabase the default database for AI-powered apps. But it is a complete machine-readable explanation of what Supabase is and what the coding agent can do with it. The tag line reads like it was written for an agent, not a person. Whether Supabase designed the agent tagline is irrelevant. Machine-readable ID is valid.
None of these companies responded. They were all answering the same thing: a visitor category that needs machine-readable identities, structured content, actionable actions, and predictable transaction flows. A growing guest category because people who follow agents love it.
What Investing in Agent Preparation Really Means
Agent readiness is not a budget line or a new team to hire. Agent readiness is a set of infrastructure decisions about how your website delivers what it offers to non-human visitors.
Can agents read your content? If your website relies on JavaScript to display its basic information, most agents see a blank page. Server-rendered HTML with semantic structure is low. Server-rendered HTML is the easiest space to cover and the one with the fastest impact.
Can agents find what you have to offer? Robots.txt that informs AI user agents, an always-up-to-date sitemap, structured data that values your businesses and relationships. These are classic web basics that are used in a separate guest class. Nothing new here. The guest is.
Can agents work? If your website sells something, can the agent complete the purchase? If your website offers a service, can an agent be persuaded? The protocol layer (UCP, MCP, WebMCP) is where this resides. Many websites are not here yet. That’s good. But the companies here are building a canal.
Moat takes it seriously before your competitors do. Shopify already enables agent toolkit access by default for all sellers. They don’t pay attention to what that means for their product data and the flow of output they are capturing agent-directed traffic. The first SaaS products with WebMCP tools discover the users discovered by the agent. Supabase is the default database for agent-based applications because its identity was machine-readable before anyone thought about it. The window is open because many websites do not start.
‘Do Something People Want’ Now Includes Agents
YC’s motto has not changed. But agents are now part of how people find what you’ve done, compare it to others, and decide whether to buy it. A social media website is a product that has a distribution problem; he doesn’t know it already exists.
Additional resources:
This post was originally published on No Hacks.
Featured image: Roman Samborskii/Shutterstock
