Paola Estefanía de Campos

Paola Estefanía de Campos is a software engineer and co-author of the Agent Auth Protocol, an open-source framework that gives AI agents distinct identities, limited permissions, and accountable access to external services. Developed with Bereket Engida within the Better Auth ecosystem, the project addresses a central security problem: agents frequently inherit human credentials when they need independently auditable, revocable authority.

Earlier in her career, de Campos held software-development roles with Reporfy and Beyomene Ventures and built experience with OAuth and OpenID Connect. Her subsequent work on agent authentication extends those foundations to software acting on someone’s behalf without impersonating them.

• Agents as independent security principals. Each agent receives its own identity and cryptographic credentials while remaining linked to the authorizing user. Services can distinguish agents, hosts, and people; audit their actions; and revoke individual agents without disabling entire accounts.

• Capability-scoped authorization. Instead of relying on broad account permissions, the protocol specification defines access around individual operations. An agent authorized to read email, for example, must receive additional permission before sending it.

• Discoverable services without shared credentials. Service directories and OpenAPI descriptions expose available operations as agent-readable capabilities, helping existing APIs participate before adopting dedicated agent-authentication standards.

• Practical oversight for autonomous agents. In a hands-on Gmail demonstration, de Campos combined separate approvals for reading and sending messages with activity logs, host policies, and revocation. She emphasizes meaningful human control without requiring approval for every low-risk action.

Her work treats secure delegation as a concrete engineering distinction: agents should receive permission to act for people without becoming indistinguishable from them.