Cisco’s “Internet of Agents” is a vision for making AI agents built by different organizations and on different platforms able to discover one another, establish identity, communicate, and be observed. Cisco links that vision to AGNTCY, an open-source project it launched in March 2025 and donated to the Linux Foundation in July 2025. It is an evolving infrastructure effort—not a finished Cisco product or a guarantee that connected agents are secure.
What is Cisco’s Internet of Agents?
Outshift by Cisco describes an Internet of Agents as infrastructure that lets agents find and work with one another regardless of who built them, which framework they use, or where they run. The vision is about interoperability across technical and organizational boundaries, rather than a single AI assistant or application.
Outshift names four infrastructure areas: discovery so agents can find one another; identity so an agent can be identified; messaging so agents can exchange information; and observability so their interactions can be monitored. These capabilities address how agents connect. They do not, by themselves, establish that an agent or its actions should be trusted. Outshift’s Internet of Agents overview sets out Cisco’s framing.
Cisco’s earlier AI infrastructure white paper describes the goal as an “open, interoperable internet for agent-agent and agent-human quantum-safe communication,” with organizational autonomy preserved. That is Cisco’s design framing; it should not be read as independent proof that every deployment is quantum-safe or secure. Cisco’s AI Infrastructure for the Agentic Era paper provides that definition.
Recommended Free Tools
#1 Best Overall
What is AGNTCY, and who runs it?
AGNTCY is the open-source project Cisco associates with its Internet of Agents vision. Outshift says it launched the project in March 2025 and donated it to the Linux Foundation in July 2025. Its intended building blocks include code, specifications, and services for discovery, identity, messaging, and observability. The Linux Foundation’s announcement likewise describes AGNTCY as foundational multi-agent infrastructure, initially open-sourced by Cisco with collaboration from LangChain and Galileo.
The governance distinction matters: Cisco originated the project and later donated it; the Linux Foundation hosts it. That history does not, on its own, establish technical neutrality, adoption, or security efficacy. See Outshift’s AGNTCY project page and the Linux Foundation’s welcome announcement.
Does agent interoperability make collaboration secure?
No. Interoperability answers whether agents using different frameworks can find and communicate with one another. Security asks whether the agents, their identities and permissions, the messages they exchange, and the actions they take are trustworthy and controlled. One does not guarantee the other.
In a January 20, 2026 post, Cisco’s Omar Santos wrote that the shift is from “AI assistants that answer” to “AI agents that act.” Cisco notes that agents can plan, call tools, invoke workflows, collaborate with other agents, and execute code—activities that expand the attack surface. Santos’s Cisco security post explains that concern.
Rank #3
In a real deployment, security also depends on controls around identity and authorization, least privilege, policy enforcement, monitoring, data governance, and the tools and actions agents can access. Discovery and messaging are infrastructure pieces; the cited project descriptions do not establish that AGNTCY alone provides every one of those safeguards or validate a particular deployment.
Cisco says it contributes to AI security through open-source tools, security frameworks, and engagement with the Coalition for Secure AI (CoSAI), OWASP, and other organizations. That describes Cisco’s activities, not an independent audit of AGNTCY or a deployment guide.
What does Cisco report about AGNTCY’s ecosystem?
Outshift’s project page reports more than 80 collaborating organizations and more than 35 repositories; the page did not give a clear as-of date for those two counts when accessed on October 4, 2026. It also says the ecosystem had nearly 150 members and six active working groups as of 2026. These are Cisco-reported figures, not independently audited counts, and their wording and dates may reflect different scopes.
The same Cisco page prints supportive statements from partner executives. Dell Technologies’ John Roese said, “Interoperability is central to Dell’s agentic AI vision. The ability of agents to work together empowers enterprises to reap the full value of AI.” Google Cloud’s Rao Surapaneni said, “Open, community-driven standards are essential for creating a diverse, interoperable agentic AI ecosystem.” These are partner perspectives, not evidence of performance or broad adoption. The figures and quotations appear on Outshift’s AGNTCY project page.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Best Value
What should organizations evaluate before relying on an agent ecosystem?
AGNTCY addresses a genuine coordination problem: agents built separately need ways to discover one another and exchange information. But an organization assessing any agent-interoperability approach should look beyond connectivity:
- Discovery and identity: How are agents located, identified, and authenticated?
- Messaging and protocols: Can agents exchange information across the frameworks and systems the organization uses?
- Permissions and data governance: What can each agent access, disclose, or change, and under whose policy?
- Observability and auditability: Can operators inspect interactions and trace consequential actions?
- Governance: Who stewards the project, and how are specifications and changes managed?
- Maturity and evidence: Is there credible evidence for the intended deployment, beyond project descriptions and reported ecosystem counts?
The available project announcements establish AGNTCY’s purpose and history, but do not provide a balanced basis for ranking it against competing frameworks or claiming comparative performance. Organizations should assess the controls and evidence for their own use case.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
