An agent’s identity does not identify the user behind it
Cloudflare argues that recognizing a hosted browsing agent does not tell a website which person authorized its actions. Many users share an operator’s infrastructure, so blocking one abusive client can also block legitimate customers. Its Trusted Agent Delegation initiative separates agent identification, user relationships and accountability, seeking interoperable mechanisms rather than assuming a CAPTCHA or permission scope answers all three.
Cloudflare · Bryan Becker · Nov 5, 2026
Why it matters
Services accepting agent traffic need an abuse boundary that can distinguish an operator from an individual account.
More context
The article explains why signed agent requests identify an agent and operator without automatically proving user authorization. TAD is a coalition examining existing protocols and bringing remaining problems to standards groups. Watch that work if building hosted agents or access policies; it is not a finished universal integration to install.
Keep in mind
This is a provider’s proposed direction. Local agents are outside its main scope, and current Cloudflare product features have separate plan and early-access requirements.
Original read