Using the NANDA Index Architecture in Practice: An Enterprise Perspective

Abstract: The proliferation of autonomous AI agents represents a paradigmatic shift from traditional web architectures toward collaborative intelligent systems requiring sophisticated mechanisms for discovery, authentication, capability verification, and secure collaboration across heterogeneous protocol environments. This paper presents a comprehensive framework addressing the fundamental infrastructure requirements for secure, trustworthy, and interoperable AI agent ecosystems. We introduce the NANDA (Networked AI Agents in a Decentralized Architecture) framework, providing global agent discovery, cryptographically verifiable capability attestation through AgentFacts, and cross-protocol interoperability across Anthropic's Modal Context Protocol (MCP), Google's Agent-to-Agent (A2A), Microsoft's NLWeb, and standard HTTPS communications. NANDA implements Zero Trust Agentic Access (ZTAA) principles, extending traditional Zero Trust Network Access (ZTNA) to address autonomous agent security challenges including capability spoofing, impersonation attacks, and sensitive data leakage. The framework defines Agent Visibility and Control (AVC) mechanisms enabling enterprise governance while maintaining operational autonomy and regulatory compliance. Our approach transforms isolated AI agents into an interconnected ecosystem of verifiable, trustworthy intelligent services, establishing foundational infrastructure for large-scale autonomous agent deployment across enterprise and consumer environments. This work addresses the critical gap between current AI agent capabilities and infrastructure requirements for secure, scalable, multi-agent collaboration, positioning the foundation for next-generation autonomous intelligent systems.

• The paper introduces the NANDA framework as a solution for secure, decentralized AI agent communication using cryptographic verification and Zero Trust principles.
• It details framework components like AgentFacts metadata and protocol adapters that enable robust agent discovery and collaboration across diverse platforms.
• The architecture demonstrates enterprise applications by automating multi-domain operations, reducing overhead while mitigating security risks such as spoofing and data leakage.

"Using the NANDA Index Architecture in Practice: An Enterprise Perspective"

Introduction

The paper presents the NANDA framework as a comprehensive infrastructure solution for AI agents, aiming to address interoperability, security, and trust challenges in autonomous agent ecosystems. The framework is designed to manage dynamic, decentralized interactions among AI agents across diverse protocol environments. Unlike traditional web architectures that align with static client-server models, NANDA introduces mechanisms for cryptographic identity verification and cross-protocol interoperability, providing a robust foundation for deploying autonomous agents in complex enterprise settings.

NANDA Framework Overview

The NANDA (Networked AI Agents in a Decentralized Architecture) framework integrates global agent discovery and securitization, combining the AgentFacts metadata and protocol interoperability adapters to enable seamless agent communication. By implementing principles of Zero Trust Agentic Access (ZTAA), the framework strengthens security against challenges like capability spoofing and impersonation attacks.

Agent visibility and control are pivotal to ensuring enterprise governance and operational compliance without hindering agent autonomy. The framework allows AI agents to discover, authenticate, and verify capabilities while managing secure agent-to-agent collaborations.

Figure 1: Building and Running an AI Agent

Architectural Components and Interoperability

To facilitate an open ecosystem, the NANDA Index operates as a discoverability layer that supports protocol interoperability among Anthropic's MCP, Google's A2A, and Microsoft's NLWeb. Agents can register and discover across diverse platforms, enabling interactions without technological constraints.

AgentFacts data provide verifiable metadata necessary for agent discovery and capability verification. Cryptographically attested claims ensure secure interactions, preventing common risks associated with decentralized agent ecosystems.

Figure 2: Agentic Architecture Breakdown

Security Concerns

Security within the NANDA framework adheres to ZTAA principles, extending traditional Zero Trust models to agentic environments. The implementation of bilateral authentication between source and destination agents, alongside capability and reputation verification, mitigates potential security breaches.

The framework addresses various security vectors, like supply chain poisoning and data leakage, ensuring safe autonomous interactions. By enforcing selective trust, based on cryptographically verified metadata, the architecture provides robust protection for agentic systems against external threats.

Figure 3: Browser-to-Webserver vs Agent-to-Agent Comparison

Enterprise Use Cases

The application of NANDA within enterprise contexts demonstrates significant efficiency improvements, particularly in complex operational scenarios like sales automation. By leveraging AI agents to automate multi-domain tasks—from customer management to feature rollouts—organizations can streamline operations while reducing overhead costs.

Figure 4: An Enterprise Web Architecture vs Agentic Architecture

Figure 5: An Enterprise Agentic Use Case for Product Operations

Conclusions

The NANDA architecture offers a foundational framework for deploying autonomous AI systems within enterprises, effectively addressing security, interoperability, and scalability issues. Its integration of Zero Trust principles into agentic environments ensures secure operations and lays the groundwork for next-generation intelligent systems. As AI agents continue to proliferate, architectures like NANDA will become increasingly essential to manage and harness their potential safely and effectively.

Figure 6: Protecting AI Agents