What’s up, Docsy? Google’s docs project joins the Linux Foundation as AI agents become readers
The New Stack Paul Sawers
Google’s Docsy docs project is moving to the Linux Foundation. It’s now being tuned for AI agents that read docs before humans do.
Based on reporting by The New Stack, Paul Sawers — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Google’s Docsy docs theme is heading to the Linux Foundation, and the timing says a lot about where technical publishing is going. The project, first announced in 2019, powers documentation for a lot of cloud-native software, and it’s now being reshaped for a world where AI tools are part of the audience too.
Erin McKean, a senior developer relations engineer at Google and a member of the Docsy steering committee, announced the move Wednesday during the Linux Foundation’s Open Source Summit Europe in Prague. Her pitch was simple: Docsy already lives inside Linux Foundation communities, so putting it under the same roof should make it easier to stay close to the projects that use it. By the end of 2024, about 2,200 projects were using Docsy, including Kubernetes, OpenTelemetry, gRPC and Jaeger.
The bigger story is the new job description for documentation. McKean says AI agents are now reading technical docs, which means the material has to be organized not just for people, but for machines that need to find and follow it cleanly. She’s blunt about the priority: if the docs help users, the route doesn’t matter much, even if that route runs through an assistant first.
Docsy has already started building for that reality. Version 0.15.0, released in May, added an experimental Markdown copy of each page and an experimental llms.txt file to index a site for AI tools. Version 0.16.0, out in July, included an upgrade guide that was written so an AI assistant could follow it with conditions, steps and checks. Then version 0.17.0, released in August, added a hidden directive at the top of pages on sites that enable llms.txt, pointing visiting agents to that index.
Next up are “AF,” or agent-friendly, documentation scores. The idea is to give maintainers a way to measure how easy their docs are for AI tools to find, navigate and use. And there’s a very old-school upside in all this too: better docs should mean fewer basic questions for maintainers to answer by hand.
My take — AI-written commentary, not fact-checked reporting
This is the rare AI feature that isn’t theatre. Making docs easier for agents is just making them easier to search, follow and trust, which is the sort of unglamorous work open source keeps skipping until the support queue starts screaming. Funny how the future of AI still looks a lot like decent documentation and fewer people asking the same question twice.
Read more about this at: The New Stack