Next.js is starting to treat AI agents as real users. That sounds odd until you think about how developers actually use these tools now. If a coding agent is going to fix a Next.js issue, it needs the same things a human would need: accurate docs, runtime errors, browser feedback, framework-specific knowledge, and a way to understand what is happening in development.
The Next.js team wrote recently about building for an agentic future, and one of the clearest points was simple: agents could not see the browser. That is a real problem. If the error is in the browser overlay but the agent only sees files and terminal output, it is missing the most important context.
That is why the direction matters. Next.js has been pushing more structured information toward the places agents can read, including terminal output, MCP integration, and project-level instructions. The newer guide also points projects toward bundled, version-matched docs and AGENTS.md files, so agents read the documentation for the version actually installed rather than guessing from whatever they find.
This is framework design now. Frameworks still need to work for humans and browsers, but they also need to work for tools that read code, inspect output, and make changes. Better logs, better errors, clearer docs, and stronger project instructions help agents, but they also help teams. That is why I think this is more than AI marketing. It is where developer experience is going.