AI agents can now talk to each other – they can’t understand what the other is trying to do. That’s the problem Cisco’s Outshift is trying to solve with a new architectural approach it calls the Internet of Cognition.
The difference is practical: Protocols like MCP and A2A let agents exchange messages and identify tools, but they don’t share intent or context. Without it, multi-agent systems burn cycles on coordination and can’t combine what they learn.
"The bottom line is, we can send messages, but the agents don’t understand each other, so there is no basis, interaction or coordination or common intention," Vijay Pandey, general manager and senior vice president at Outshift, told VentureBeat.
Practical Effects:
Consider scheduling a specialist appointment for a patient. With MCP alone, a symptom assessment agent sends a diagnosis code to a scheduling agent, who locates available appointments. An insurance agent verifies coverage. A pharmacy agent checks for medication availability.
Each agent accomplishes its task, but none of them reason together about the patient’s needs. The pharmacy agent may recommend a medication that conflicts with the patient’s history – symptoms the agent has the information for but did not proceed with because "Potential Drug Interactions" Was not within its scope. The scheduling agent books the nearest available appointment, without knowing that the insurance agent has found better coverage at a different facility.
They’re connected, but they’re not aligned on the goal: find the right care for this patient’s specific situation.
Current protocols handle the mechanics of agent communication – MCP, A2A, and Outshift’s AGNTCY, which it donated to the Linux Foundation, lets agents find devices and exchange messages. But they work on what is called Pandey "Connectivity and identity layer." They handle syntax, not semantics.
The missing piece is shared context and intent. The agent completing a task knows what it is doing and why it is doing it, but that logic is not transmitted when it delegates the task to another agent. Each agent interprets goals independently, meaning coordination requires constant clarification and learned insights remain tacit.
According to Outshift, for agents to move from communication to collaboration, they need to share three things: pattern recognition in the dataset, causal relationships between tasks, and clear goal states.
"Without shared intent and shared context, AI agents remain semantically isolated. They are individually capable, but goals are interpreted differently; Burns the coordination circle, and compounds nothing. One agent learns something valuable, but the rest of the multi-agent-human organization is still starting from scratch," OutShift said in a paper. Outshift said the industry needs it "Open, interoperable, enterprise-grade agentic systems that collaborate meaningfully" and proposes a new architecture which he calls "internet of perception," Where multi-agent environments operate within a shared system.
The proposed architecture introduces three layers:
Cognition State Protocol: A semantic layer that sits on top of a message-passing protocol. Agents share not only data but also intent – what they are trying to achieve and why. This lets agents focus on goals before acting rather than providing explanations after the fact.
Feeling Fabric: Infrastructure for creating and maintaining shared context. Think of it as distributed working memory: context graphs that persist across agent interactions, with policy controls for what is shared and who can access it. What can system designers define "common sense" Their use case seems to be like this.
Cognition Engine: Two types of capacity. Accelerators allow agents to learn insights and blended learning – one agent’s discoveries become available to others solving related problems. Guardrails enforce compliance boundaries so that shared logic does not violate regulatory or policy constraints.
OutShift positioned the framework as a call to action rather than a finished product. The company is working on implementation, but stressed that semantic agent collaboration will require industry-wide coordination – just as early Internet protocols required buy-in to become standard. Outshift is in the process of writing code, publishing specifications, and releasing research around the Internet of Cognition. It is expected to get a demo of the protocol soon. Noah Goodman, co-founder of frontier AI company Humans and professor of computer science at Stanford, said during VentureBeat’s AI Impact event held in San Francisco that innovation happens when "Other humans figure out which humans to pay attention to." The same dynamic applies to agent systems: as individual agents learn, the value multiplies when other agents can recognize and leverage that knowledge. Now the practical question for teams deploying multi-agent systems: Are your agents just connected, or are they actually working toward the same goal?
<a href