A standard, open framework for building AI agents is coming from Cisco, LangChain and Galileo

A goal Agent Future It is aimed at AI agents from different organizations, and can talk to each other freely and seamlessly. But to this point, interoperability is required, and these agents may be built using different LLMs, data frameworks, and code.

To achieve interoperability, developers of these agents must be on how Communication with each other. This is a challenging task.

A group of companies including Cisco,,,,, Langchain,,,,, Llamaindex,,,,, Galileo and collectnow created Agntcy, an open source collective with the goal of creating an industry-standard agent interoperability language. Agntcy is designed to make it easy for any AI agent to communicate and exchange data with another agent.

Unity AI Agent

“Just like the cloud and the internet appear and accelerates applications and all social interactions around the world, we want to build an Internet that promotes all human works around the world,” said Vijoy Pandey, head of OutShift at Cisco, in an interview with VentureBeat.

Pandey compared Agntcy to Transmission Control Protocol/Internet Protocol (TCP/IP) and Domain Name System (DNS)This helps organize the Internet and allows interconnection between different computer systems.

“The way we think about this is that the original Internet allows humans, servers and web farms to come together,” he said. “This is the Internet of Nek, and the only way is to make it open and interoperable.”

Cisco, Lamban and Galileo will act as core maintainers of Agntcy with Glean and Llamaindex as contributors. However, this structure may change as the collective adds more members.

Standardize a rapidly developing industry

Artificial intelligence agents cannot be islands. To reach their full potential, they must be able to communicate with other agencies outside of the corporate network. This is the source of interoperability.

Setting standards in traditional industries is challenging enough. For a technology like AI, it becomes even more difficult, where upgrades and model changes occur every few months. However, this is not the first time that standards for generating AI have been proposed.

Langchain, one of the core members of Agntcy, has his own agreement to work with other frameworks other than Mrs. Langchain. this Agent agreement, Launched last November, Langchain agents are allowed to talk to agents created using Autogen, Crewai or any other framework.

At the same time, humans declare November Model Context Protocol (MCP). The protocol is designed to standardize how models and AI tools connect to data sources. But while many developers accept MCP, it is not exactly the standard yet.

Yash Sheth, co-founder of AI evaluation platform Galileo, said standardization is “critical.”

“In fact, standardization is needed, and it will drive the speed of adoption of agent adoption. Today, teams are building silos and have to figure out how to develop their own infrastructure components from scratch,” Sheth said in an email. “The standardization of multi-agent systems can only occur when these agents powered by non-deterministic models have strong anchor points in measuring and reporting their performance, accuracy and reliability.”

Sheth acknowledged that making AI agents interoperable can be complicated. Agntcy “hopes to encourage developers to extend these specifications, APIs, and tools to suit their needs, rather than reinventing the wheel, which is crucial to achieving standardization.”

Langchain CEO Harrison Chase said in a separate conversation that creating standards is not impossible, especially since building an agent would be easier.

“Building agents are already possible and are being done. Replit, Klarna, LinkedIn, Uber, AppFolio and many others have done this. Agents are not 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. That’s what the standard proxy protocol will help achieve,” Chase said.

Platform and one language at a time

Pandey envisions Agntcy as more than just a set of code for an agent. It will also allow customers to discover agents of different developers running the AGNTCY standard.

“Customers can stitch all of these agents on the Agntcy platform to discover, make up, deploy and evaluate when building workflows,” Pandy said.

Agntcy still needs to recruit more AI players to add new agents to the platform and become the driving force for the standard. After all, to be an industry standard, large-scale adoption is required to prevent too many competitive standards from being established.

That’s where projects like Agntcy face tough battles. Pandy said the collective has been talking to many other industry players who want to get as many perspectives as they develop the platform. That will take time.

At the same time, enterprises continue to experiment and even deploy AI agents. Maybe in the future, these can all be talked to each other.