TLDR: MIT’s Project NANDA, or Network of AI Agents and Decentralized Architecture, is developing an open, decentralized infrastructure to enable seamless discovery, communication, and collaboration among billions of AI agents. Building on protocols from Anthropic and Google, NANDA aims to create a robust, secure, and interoperable ecosystem for autonomous AI systems, addressing challenges like trust, privacy, and scalability.
In a significant move to reshape the future of artificial intelligence, MIT’s Project NANDA (Network of AI Agents and Decentralized Architecture) is spearheading the creation of an open, decentralized infrastructure for AI agents. This ambitious initiative seeks to move beyond the current landscape of proprietary AI platforms, envisioning a true ‘Internet of AI Agents’ where billions of specialized AI entities can discover, interact, and collaborate autonomously across diverse networks.
The motivation behind Project NANDA stems from the observation that while 2025 has seen a surge in AI agent development, most attention has been on frameworks for building these agents, with less focus on their distribution and interoperability. The project aims to address this gap by providing foundational building blocks for a democratic and widespread agent ecosystem.
At its core, NANDA’s technical architecture is designed to facilitate cross-platform and cross-vendor interactions. It builds upon existing protocols such as Anthropic’s Model Context Protocol (MCP) and Google’s Agent-to-Agent (A2A) Protocol. Key components of this infrastructure include:
The NANDA Index: A globally distributed reference index that maps an agent’s handle to a verified metadata file or endpoint, akin to a decentralized DNS for AI agents.
Decentralized Registry System: This system functions as a discovery and authentication mechanism, enabling verifiable interactions across the network.
Advanced Communication Protocols: These protocols provide the bedrock for AI communication, allowing agents to find each other and query distributed knowledge across networks.
Security and Verification: NANDA incorporates secure verification protocols to ensure trustworthy agent interactions, including verifiable accountability, native identity, traceability, and behavioral records.
The vision articulated by the MIT Media Lab is one where AI agents can seamlessly perform discrete functions, navigate autonomously, socialize, learn, earn, and transact on behalf of users. This collaborative environment is expected to include federated registries and open identity and trust layers, fostering an agent ecosystem that is both robust and flexible.
Prof. Ramesh Raskar’s leadership has guided this initiative for ten years, bringing together a network of 18 leading research institutions to advance networked agent intelligence. The project emphasizes open-source development and encourages participation in working and interest groups to shape specifications and guidelines for the agentic web.
However, the path to this decentralized future is not without its challenges. Project NANDA is actively addressing critical issues such as privacy, the development of data markets, verifiability, authentication, and the creation of intuitive UI/UX for human-agent interaction. The goal is to make onboarding to such an ecosystem easy, democratic, and widespread.
Also Read:
- INFINIT and Google Forge Alliance to Develop Global Intelligent Agent Financial Infrastructure
- AI’s Transformative Impact: The Top Trends of 2025 Reshaping Daily Life
Major milestones for Project NANDA include ongoing events at the MIT Media Lab and continuous development of the foundational protocols that will underpin the next generation of distributed artificial intelligence. The project aims to position participants at the forefront of creating this ‘Internet of AI Agents,’ accelerating the adoption of new AI products and tools.
