Outshift lays out coordination layer for multi-agent AI

Outshift by Cisco is positioning a semantic coordination layer, described as the “Internet of Cognition,” as the missing piece for multi-agent AI systems that can work across domains without constant human stitching. Vijoy Pandey says today’s agents can exchange data but lack shared intent, context, and reasoning, leaving systems such as healthcare workflows dependent on humans to coordinate decisions.

The approach builds on AGNTCY, an open-source connectivity layer now under the Linux Foundation, designed to let agents across systems, companies, and platforms discover one another, prove identity, and exchange messages through standardized protocols. One cited study found a failure rate of between 41% and around 87% in evaluated multi-agent systems, underscoring Pandey’s view that the challenge is architectural rather than a prompting problem.

Outshift’s stack includes Mycelium for goal coordination, a cognition fabric for shared institutional memory, and guardrail technologies for security, cost, and compliance. Internal testing found unstructured groups reached a decision about a third of the time across 14 scenarios, while a coordination protocol raised that to 93%. CASA, another open-source reference implementation, continuously checks whether an agent’s tool requests still match the user’s original goal.

Pandey advises enterprises to start with a cross-functional workflow that spans three or four teams and still relies on human authorization. The goal is to measure whether one agent’s insight improves another agent’s performance as organizations add horizontal coordination on top of larger models.