The Agent Access Index.
A pilot dataset of what agents encounter before fetching. It starts with first-party routing decisions and an explicit methodology. It is not yet a representative survey of the web.
Last 30 days
Loading observations…
* Anonymous client keys rotate daily and cannot be summed into unique people or agents. Registered clients are operator-provisioned integrations. Neither proves autonomous usage.
Returning registered integrations
What would make this useful?
With enough opted-in external integrations, report restriction frequency, advertised interface coverage, evidence freshness, and task outcomes by cohort. Suppress small samples, disclose selection bias, and keep client reports separate from independently measured results.
What this is today
Aggregate observations from this service only. Demo traffic is separate. No claim of customers, revenue, market share, or web-wide reliability. A recommendation succeeding at preflight does not prove that an agent completed its intended task.