Agent traffic
Agent traffic is the portion of requests to a website or API that originates from AI systems rather than humans in browsers — crawlers gathering training or retrieval data, assistants fetching pages to answer questions, and agents reading documentation while executing tasks. It is typically identified by declared user agents such as GPTBot, ClaudeBot, and Google-Extended, though not every agent self-identifies.
Agent traffic behaves differently from human traffic. It fetches over plain HTTP and generally executes no JavaScript, so a client-rendered page can return an empty shell to an agent that a human browser would render fully. It views no ads and fills no forms, and its access is partly a policy decision — robots.txt directives addressed to specific AI crawlers determine which agents a site admits.
Producers watch agent traffic because it is rising in their analytics while the agent's experience remains invisible. The log line shows a fetch; it does not show whether the agent found the reference, parsed the spec, or gave up on an empty response. Each of those fetches is a machine reading the docs with intent — often on its way to building, or abandoning, an integration.
Discry measures what agent traffic encounters on arrival: whether a plain HTTP fetch of the API's public documentation surface yields a machine-legible entry point, and whether models can operate the API from what that fetch returns. Traffic volume itself sits outside the measurement — the Discry Score characterizes the readiness of the surface those requests hit, before a single API call is made.
What does an AI agent make of your API?
Find out in about a minute — no signup.
Discry your API — free