Cloudflare, Tools for Humanity propose agent identity framework
Cloudflare and Tools for Humanity announced the Trusted Agent Delegation (TAD) coalition on November 5, 2026, introducing a framework designed to let websites authenticate AI agents and assign reputation scores without blocking legitimate bot traffic or requiring every user to prove their identity Cloudflare.
The Problem: Distinguishing Agents from Abuse
As AI agents proliferate across the web—automating research, data collection, testing, and customer service—websites face a dilemma: how do you let good agents in while keeping out fraud, scraping, and malicious automation? Traditional solutions are blunt: block all non-human traffic, or require every user to prove they are human. Neither works at scale for an agent economy.
The TAD proposal addresses this tension through a framework built on three core functions: identity (knowing who the agent and its operator are), reputation (tracking their past behavior), and accountability (creating consequences for abuse) Biometric Update.
Four Tiers of User Identification
Instead of a binary "human or bot" check, TAD proposes four levels of operator identification, each suited to different risk profiles:
• Nothing: No user identifier supplied; the site evaluates only the agent itself.
• Session: Requests tied to a single agent session.
• Account: Requests linked across sessions through a stable operator or third-party identifier.
• Unique: A persistent third-party identifier for higher-risk actions—such as a government-issued ID or a proof-of-human credential like World ID.
This tiered approach lets website operators choose the verification level appropriate to their threat model. Low-risk read operations might require only agent reputation; sensitive actions like account creation or financial transactions could demand a unique human credential.
Technical Implementation: Cryptographic Signatures
Cloudflare's Web Bot Auth method enables agents to cryptographically sign requests, cryptographically proving the identity of the agent and its operator to receiving websites Yahoo Cybersecurity. This allows verification without forcing the user's identity into the public record—the operator proves who they are to the website, not to the internet.
A Coalition Framework, Not Yet a Standard
The TAD coalition brings ecosystem representatives together to examine existing protocols and, where necessary, bring better-defined problems to the IETF for standards work. The current announcement is a framework for discussion and collaborative standard-setting, not a completed formal specification Cloudflare.
This positions TAD as an open-source governance model for the agent economy—establishing shared language and architecture before fragmentation locks in incompatible approaches across platforms.
