Most agent-identity proposals solve one layer at a time — a DID scheme here, a delegation format there. LOKA (Layered Orchestration for Knowledgeful Agents), a protocol out of Carnegie Mellon, tries to bundle three layers into a single stack: identity, communication, and ethics. The identity layer is a Universal Agent Identity Layer (UAIL). A recent survey of AI agent protocols describes it plainly: LOKA proposes a Universal Agent Identity Layer to assign unique, verifiable identities to AI agents, facilitating secure authentication, accountability, and interoperability, and it builds intent-centric communication on top of that so agents can coordinate semantically rather than just exchanging opaque API calls. The identity itself is meant to be self-sovereign rather than issued by a platform. Based on Self-Sovereign Identity principles, the layer assigns each AI agent a unique, verifiable, and portable identity, and the agent — not a central authority — controls this identity, establishing the foundation for accountability. That's the part worth pausing on: accountability here isn't a logging feature bolted on afterward, it's supposed to fall out of the fact that the agent's identity is cryptographically its own and can't be silently reassigned or revoked by whoever happens to run the platform. Above identity and communication sits what LOKA calls a Decentralized Ethical Consensus Protocol — a mechanism for agents to weigh actions against shared ethical baselines that can still flex by jurisdiction. The protocol's own framing names the gap directly: AI agents often operate within siloed systems, lacking a common protocol for communication, ethical reasoning, and compliance with jurisdictional regulations, and this fragmentation poses risks including interoperability issues, ethical misalignment, and accountability gaps. Whether the ethics layer is workable is a separate, harder question than the identity layer — encoding 'shared ethical baselines' into consensus logic is a much fuzzier problem than issuing a DID. But the framing is useful regardless of whether DECP survives contact with real deployments: it treats accountability as something that has to be designed in at the identity layer, not inferred later from behavior logs. Later work building on LOKA — including efforts embedding verifiable-credential exchange directly into agent-to-agent protocols — has kept the identity piece and mostly left the ethics-consensus piece for future research, which suggests the accountability layer is the part of LOKA's stack that's actually proving portable.