Join our daily and weekly newsletters for the latest updates and the exclusive content on AI coverage. Learn more
A goal for a agentic future is for AI agents of different organizations to speak freely and transparently. But getting to this point requires interoperability, and these agents may have been built with different LLM, data frames and code.
To achieve interoperability, the developers of these agents must agree on how they can communicate. It is a difficult task.
A group of companies, including Cisco,, Lubricole,, Llamandex,, Galileo And Gleanhave now created Agntcy, an open source collective in order to create a standard agent interoperability language in the industry. Agntcy aims to facilitate any AI agent to communicate and exchange data with another.
Unite AI agents
“Just like when the Cloud and the Internet have appeared and accelerated applications and all social interactions worldwide, we want to build the Internet of agents who accelerate all human work on a global scale,” said Vijoy Pandey, head of Venturebeat.
Pandey compared Agntcy to the advent of Transmission control protocol / Internet protocol (TCP / IP) and the domain name system (DNS)This helped organize the Internet and allowed interconnections between different computer systems.
“The way we think of this problem is that the original Internet has enabled humans and servers and web farms to come together,” he said. “This is the internet of agents, and the only way to do so is to open it and interoperable.”
Cisco, Langchain and Galileo will act as the basic officials of Agntcy, with Glean and Llamaindex as contributors. However, this structure can change because the collective adds more members.
Standardize a quick industry
AI agents cannot be islands. To reach their full potential, they must be able to communicate with other agents who are outside the network of a company. This is where interoperability comes into play.
Making standards in traditional industries is difficult enough; It becomes even more difficult for technology like AI, where upgrades and model changes occur every few months. However, this is not the first time that a standard has been offered for generative AI.
Langchain, one of Agntcy’s basic members, has his own protocol to work with agents built on frames other than Langchain. THE Agent protocol, Launched in November of last year, allows Langchain agents to speak to agents created with Autogen, Crewai or any other executive.
Meanwhile, Anthropic announced his Model context protocol (MCP) in November. This protocol aims to normalize how AI models and tools connect to data sources. But while many developers have adopted MCP, it is not yet exactly a standard.
Yash Sheth, co-founder of the AI Galileo assessment platform, said that normalization “is critical”.
“Normalization is necessary, in fact, it will result in increased speed for agentic adoption. Today, the teams are built in silos, having to find out how to develop their own infrastructure components from scratch, “said Sheth in an email. “The standardization of multi-agent systems can only occur if these agents supplied by non-deterministic models have a strong anchor insofar as and the declaration of their performance, their precision and their reliability.”
Sheth admits that the manufacture of interoperable AI agents can be complex. Agntcy “wants to encourage developers to extend these specifications, APIs and tools to meet their needs instead of reinventing the wheel, which will be crucial to reach standardization”.
The CEO of Langchain, Harrison Chase, said in a separate conversation that the creation of a standard is not impossible, especially now that it is easier to build the agents themselves.
“Construction agents are already possible and the fact. Folds, Klarna, LinkedIn, Uber, Appfolio and many others have already done so. Agents are not one thing about the future, they are now. Now that we know how to build agents, the next step is to allow them to connect to each other. This is what a standard agent protocol will help to allow, “said Chase.
A platform and a language at a time
Pandey envisages Agntcy as more than a simple set of codes for agents. It will also allow customers to discover agents from different developers who manage the Agntcy standard.
“Customers can assemble all these agents on the AGNTCY platform so that they can discover, compose, deploy and assess when they build their workflows,” said Pandey.
Agntcy must still recruit more AI players to add new agents to the platform and take momentum as a standard. After all, for something to become an industry standard, there must be mass adoption, to prevent the establishment of too many competing standards.
This is where projects like Agntcy face a difficult battle. Pandey said that the collective had spoken with many other industry players, and that they want to get as many points of view as possible during the development of the platform. It will take time.
In the meantime, companies continue to experiment and even deploy AI agents. Perhaps in the future, they will all be able to talk to each other.
