Cloudflare, the internet infrastructure and security company, has launched tools that let websites track and control personal AI agents.
These are software assistants that browse the web on behalf of individual users, rather than the automated crawlers that scrape sites in bulk.
Cloudflare's Security Analytics dashboard now includes a "Personal Assistants" category alongside its existing types of bot traffic.
It shows site owners which agents are visiting and how many requests each one makes.
Owners can also filter that traffic by the pages requested, the responses returned and where the traffic comes from, to see exactly what agents are doing.
More choice
Until now, sites could only allow or block all agents with a single switch.
They can now choose to allow individual agents, block them, or block them only on pages that carry adverts.
Any agent not set individually follows the site's general policy.
Customers using Account Abuse Protection, which is in early access, can also see which agents are linked to user logins and sign-ups.
If an account looks suspicious, fraud teams can block that account alone using a firewall rule, without shutting the agent out for everyone.
Cloudflare said this meant analysts could stop an attacker using an agent on an account without blocking it for everyone else who relies on it.
Rapid growth
The launch follows a surge in agent traffic.
Daily requests from AI agents on Cloudflare's network grew by more than 1,700% over the past year.
Overall traffic rose from about 63 million requests per second at the end of 2024 to roughly 115 million, peaking at more than 150 million.
Cloudflare tracks AI training crawlers separately.
Read more about Cloud Infrastructure
