The founder of Inth, a startup that builds developer tools, made this point in the AI Engineer video series: more and more often, software is chosen and installed by an AI agent, not by a person browsing the product's website. That means companies need to rethink how they acquire customers. Marketing is moving from HTML pages to instructions embedded in code as Markdown files. It's a new inbound channel, and the customers are the agents themselves.
What this means:
- Agents have taken over the sales channel. Instead of traditional web traffic metrics, you need to look at referrals from systems like Claude, ChatGPT, Codex and Gemini, and track install metrics too.
- Marketing content is source code. Product information has to live inside Node modules, code repositories and local Markdown files. Help centers on websites are losing weight: it's easier for an agent to read a file sitting next to the code.
- Savings on compute. Having AI read local Markdown files instead of crawling the web cuts the compute bill in half. And in 2026 Cloudflare noted that non-human traffic already makes up more than half of all internet traffic.
The first thing I would check in my own projects: how we currently package information for AI agents that might want to use our solutions. This isn't about convenience, it's about cost and speed. How can your libraries and utilities be "found" by agents instead of people? How can they be installed and configured with no human involved and no long parsing of a complex web interface?
For anyone building agentic pipelines or developing RAG systems, this is a signal to act: start packaging your documentation and setup instructions in machine-readable formats so agents can integrate with your software on their own. For product teams working on developer tools, it's a shift in focus: your new customers are code and AI.