AGNTCY: Building the Future of Multi-Agentic Systems

The true ROI of Artificial Intelligence comes through automation. As organizations increasingly adopt AI, we've naturally gravitated toward agentic architectures to apply this technology across unprecedented use cases and workflows. However, this comes with its own skepticism and predictions. As we all rush to build the agentic future, there is a strong need for standardization to truly make the multi-agentic system a reality.

Our Mission

To empower every AI developer to seamlessly Build, Ship and Scale, AI-powered applications in a trustworthy and reliable manner.

Challenges

While singular agents perform small tasks, we can only achieve scalable automation when these agents can collaborate with other agents to solve complex, multi-faceted problems.

With the onslaught of AI-powered development where every application—from web services to consumer applications, data pipelines, and control systems—is becoming “agentic” in nature, there are significant infrastructure gaps preventing this future from materializing:

1. Lack of Standardization:

AI developers are building agentic applications in silos, using a variety of custom frameworks, design patterns, APIs, and architectures, leveraging a diverse technology stack. While this is still true of traditional software development, there is an increased need for a standard way for developers to Define, Discover, and Deploy their agents, as well as a standard communication protocol for agents-to-agent collaboration.

2. Lack of Trust and Reliability:

With agents fundamentally powered by a non-deterministic technology that makes API calls and modifies real-world systems in irreversible ways, ensuring that the behavior of these agents can be measured/quantified by metrics is key. While measurement is the first problem we need to solve for, Agents also need to report their active behavior for better transparency, auditability, and observability.

3. Lack of Key Infrastructure:

In a Multi-Agent System, agents being built in a bespoke manner need critical infrastructure components such as Deployment at scale, Authentication / Access management, State management (ephemeral and serialized), Encryption / Security, and Universal access to Tools and Capabilities.

AGNTCY: An Open Source Collective to Overcome These Challenges.

Galileo, Langchain, and Cisco have come together to begin building towards a truly Multi-Agentic future, with Glean and LlamaIndex joining as supporting members.

AGNTCY is an open collective that aims to build and adopt critical open-source infrastructure components to address the key challenges noted above. We invite all AI development teams and industry partners to join us in advancing AGNTCY by adopting and extending these components/specs to serve their specific domains and use cases.

Our Roadmap: Four Critical Phases

We've identified four crucial phases needed to deliver effective multi-agent workflows:

1. DISCOVER: Find and evaluate the best agents for the job

2. COMPOSE: Connect agents into effective workflows across frameworks and vendors

3. DEPLOY: Run multi-agent systems at scale, securely

4. EVALUATE: Monitor performance and improve efficiency and efficacy over time

Getting Started With AGNTCY

We’re systematically tackling this entire infrastructure to bring the Internet of Agents to life. Start by exploring the following specs (with code dropping soon):

DISCOVER

COMPOSE

What could these look like in action? Our goal is seamless, end-to-end agent-agent collaboration: A developer can use the Agent Directory to find agents that match their requirements (as defined by OASF), then use the Agent Connect Protocol to enable communication regardless of which frameworks the agents were built with.

Expanding The Collective

LangChain, Galileo, and Cisco have come together to start this journey, but building the Internet of Agents requires broader collaboration. Help us unlock agentic capabilities through open source by joining AGNTCY. There are countless opportunities for members to contribute expertise. We're currently exploring numerous contribution opportunities:

• Agent orchestration and workflow management

• Agent performance monitoring and evaluation

• Domain-specific agent frameworks and OASF extensions

• AI infrastructure and hardware acceleration

• Sample new applications and uses cases

Join Us

The transition to agentic-first computing at scale isn’t a matter of if but of when. Getting the proper infrastructure in place now will determine just how transformative these systems will be in the years ahead.

Visit agntcy.org to explore our specifications. Code will be available at the end of March. If you’re interested in becoming a member, email us at hello@agntcy.org (we do have some humans behind the inbox).

Together, we will build a global standard that enables all of us to Build, Ship, and Scale faster.