Photo credit: venturebeat.com
Stay updated with the latest advancements and insights in the AI industry by subscribing to our newsletters.
A significant aspiration for the future of artificial intelligence is the seamless interaction among AI agents developed by different organizations. Realizing this vision hinges on achieving interoperability, as these agents may utilize distinct language models (LLMs), data structures, and coding practices.
For interoperability to be successful, developers must establish a common framework for communication among their AI agents. This presents a formidable challenge.
A coalition of companies, including Cisco, LangChain, LlamaIndex, Galileo, and Glean, has formed AGNTCY, an open-source initiative aimed at establishing a universal interoperability language for AI agents. The objective of AGNTCY is to facilitate straightforward communication and data exchange between diverse AI systems.
Uniting AI Agents
“Much like the emergence of the cloud and internet transformed applications and global interactions, our goal is to create the Internet of Agents that will enhance all human endeavors on a worldwide scale,” stated Vijoy Pandey, head of Outshift by Cisco, during an interview with VentureBeat.
Pandey drew parallels between AGNTCY and the early standards such as Transmission Control Protocol/Internet Protocol (TCP/IP) and the domain name system (DNS), which organized the internet and enabled networking among various computer systems.
“Our perspective is that the original internet unified humans, servers, and web farms,” he remarked. “This is what we envision for the Internet of Agents—an ecosystem that thrives on openness and interoperability.”
Cisco, LangChain, and Galileo will be the primary maintainers of AGNTCY, with Glean and LlamaIndex contributing as members. This structure may evolve as more organizations join the collective.
Standardizing a Fast-Moving Industry
For AI agents to reach their full potential, they must be able to communicate beyond their respective enterprises. This necessity underscores the importance of interoperability.
Establishing standards in traditional sectors is difficult; for fast-evolving technologies such as AI, where updates and new models are frequent, the challenge is even greater. However, this is not the first attempt to set standards within the generative AI landscape.
LangChain, a core member of AGNTCY, has introduced its own protocol for interoperability among agents built on different frameworks. The Agent Protocol, unveiled last November, enables LangChain agents to communicate with those developed using AutoGen, CrewAI, or other frameworks.
Similarly, Anthropic introduced its Model Context Protocol (MCP) around the same time, aiming to standardize model interactions with data sources. Although many developers have shown interest in MCP, it has not yet achieved the status of a formal standard.
Yash Sheth, co-founder of AI evaluation platform Galileo, emphasized the necessity of standardization. “It’s essential for speeding up agentic adoption. Currently, teams are often confined to their silos, forced to build infrastructure components independently,” Sheth elaborated in an email. “Standardization of multi-agent systems can only progress if these agents, driven by non-deterministic models, have a solid foundation in assessing and reporting their performance, accuracy, and reliability.”
Sheth acknowledges that enabling interoperability among AI agents is complex. AGNTCY aims to inspire developers to adapt specifications, APIs, and tools to fit their needs rather than starting from scratch, which is crucial for achieving broader standardization.
LangChain CEO Harrison Chase expressed optimism about establishing a standard, especially as the mechanisms for creating agents become increasingly accessible. “Today, constructing agents is very much feasible. Organizations like Replit, Klarna, LinkedIn, and others are actively engaged in this. Agents are not a distant future—they are a current reality. The next step is enabling these agents to connect, and that is where a standard agent protocol will be transformative,” Chase asserted.
A Platform and a Language All at Once
Pandey envisions AGNTCY evolving beyond a mere specification for agents. It will also provide a platform for customers to explore agents from various developers that comply with the AGNTCY standard.
“Customers will be able to integrate multiple agents through the AGNTCY platform, thereby discovering, composing, deploying, and assessing as they develop their workflows,” he remarked.
The initiative faces the challenge of recruiting more participants in the AI landscape to enhance the platform with additional agents, which is vital for achieving widespread acceptance and preventing the emergence of competing standards.
As it stands, AGNTCY is in discussions with numerous industry players to gather a wide range of perspectives during the platform’s development, a process that will require time and cooperation.
Meanwhile, enterprises are actively experimenting with and deploying AI agents, paving the way for a future where these diverse systems can communicate effectively.
Source
venturebeat.com