Photo credit: venturebeat.com
Agentic interoperability is on the rise, although many organizations are still exploring which interoperability protocols to implement as the landscape continues to evolve.
A team of researchers from Carnegie Mellon University recently introduced a novel interoperability protocol aimed at governing the identity, accountability, and ethical considerations of autonomous AI agents. This initiative, named Layered Orchestration for Knowledgeful Agents (LOKA), joins a list of suggested frameworks that includes Google’s Agent2Agent (A2A) and Anthropic’s Model Context Protocol (MCP).
In a detailed research paper, the team emphasized the growing presence of AI agents and the pressing need for structured governance.
“With their increasing prevalence, establishing a standardized framework for AI agent interactions has become critical,” the researchers stated. “Currently, many AI agents function in isolated environments, lacking a universal protocol for communication, ethical considerations, and compliance with regulatory requirements. This disjointed operation creates risks, including interoperability challenges, ethical discrepancies, and accountability issues.”
To remedy these challenges, the researchers advocate for LOKA, intended as an open-source solution by providing agents with a means to establish their identity, exchange ethically-informed messages, integrate accountability, and implement ethical governance throughout their decision-making processes.
At the core of LOKA is a Universal Agent Identity Layer, which provides agents with a distinct and verifiable identity.
“We see LOKA as not only a foundational architecture but also a call to reassess critical components—such as identity, intent, trust, and ethical consensus—that should guide agent interactions. As the role of AI agents grows, it’s essential to evaluate whether our current infrastructure can effectively support this transition,” Rajesh Ranjan, one of the researchers, shared with VentureBeat.
The Layers of LOKA
LOKA operates as a multilayered structure. The initial layer focuses on identity, which defines what an agent is. This includes a decentralized identifier that serves as a “unique, cryptographically verifiable ID,” enabling users and fellow agents to confirm the agent’s identity.
The subsequent layer pertains to communication, where the agent clarifies its intent and the specific task it aims to fulfill. Following this is the ethics layer, accompanied by a security layer.
The ethics layer delineates agent behavior. It encompasses “a flexible yet robust ethical decision-making framework” that allows agents to adjust their actions according to different ethical standards, depending on their operational context. This LOKA protocol employs collective decision-making models, enabling agents within the framework to evaluate their course of action and determine whether it aligns with responsible AI standards.
In addition, the security layer employs what the researchers term “quantum-resilient cryptography.”
Distinct Features of LOKA
The researchers noted that LOKA differentiates itself by enabling essential information sharing among agents, thereby allowing them to function autonomously across varied systems.
This protocol could provide enterprises with a means to ensure the safety of their deployed agents, while also offering a traceable framework to comprehend the decision-making process of these agents. Concerns often loomed for organizations regarding the possibility of an agent mistakenly accessing confidential data from another system.
Ranjan mentioned that the framework “underscores the need to clearly define who the agents are, how they make decisions, and how they are held accountable.”
“Our vision is to highlight the pivotal questions that often get overlooked during the rapid scaling of AI agents: How can we build ecosystems in which these agents can be trusted, held accountable, and interact ethically across various systems?” Ranjan articulated.
While LOKA will need to compete with emerging agentic protocols—like MCP and A2A, which have garnered interest due to their backing by well-known organizations—its independent progress has excited the researchers. Ranjan noted that they have received “very encouraging and exciting feedback” from peers and institutions anticipating further developments in the LOKA project.
Source
venturebeat.com