Agents as users
Machine-readable context
Machine-readable context is documentation and structure an agent can consume directly, including markdown documentation, copy-pasteable examples, an llms.txt index and semantic markup.
Control sits with the agent
When to use it
Any product an agent might operate on a customer’s behalf, which now includes most commerce and booking flows.
What it prevents
Agents failing on interfaces built for visual inference, where a control a person reads at a glance is read as something else entirely.
What an agent findsbuilt for reading
llms.txtAn index of what exists and where
MarkdownDocumentation without a layout to infer
ExamplesCopyable, and correct as written
Icon onlyThe one thing it will read wrong
Who ships it
- The agent experience community site initiated by Netlify agentexperience.ax
Related patterns: Skills · The protocol stack