NANDA (Networked Agents and Decentralized AI) is an ambitious MIT initiative developing the foundational infrastructure for a true "Internet of AI Agents." Just as the early web required DNS and HTTP protocols to scale, NANDA is creating the essential building blocks to support trillions of autonomous AI agents that can discover, verify, and collaborate in real-time across decentralized networks.
Vision and Motivation
The project envisions billions of specialized AI agents collaborating across a decentralized architecture, where each agent performs discrete functions while communicating seamlessly, navigating autonomously, socializing, learning, earning and transacting on our behalf. At the heart of the project lies a fundamental question: Can we upgrade the current internet—or do we need a new one entirely?
Technical Architecture
NANDA builds on Anthropic's Model Context Protocol (MCP) and Google's Agent-to-Agent (A2A) to create comprehensive distributed agent intelligence infrastructure. The system introduces several critical components:
1. Decentralized Registry System: Functions like DNS for agents, enabling discovery, authentication, and verifiable interaction across the network.
2. Advanced Communication Protocols: Agent-to-agent protocols providing the foundation for AI communication, with mechanisms for agents to find each other in the network and systems for querying distributed knowledge across agent networks.
3. Security and Verification: Secure verification protocols for trustworthy agent interactions, including verifiable agent-to-agent exchange accountability and native identity, traceability, and behavioral records.
4. Developer Ecosystem: Services for developers to build on the NANDA ecosystem, including registry of agents, resources and tools, database of agentic interactions, and third-party integrations with client and server SDKs.
Research Impact and Collaboration
The project represents a network of leading institutions collaborating to advance the frontiers of networked agent intelligence, with 18 leading research institutions pioneering the future of AI. This initiative, ten years in development under Prof. Ramesh Raskar's leadership, introduces the critical infrastructure needed to scale distributed agent intelligence.
Community and Engagement
The project maintains an active ecosystem through multiple channels:
- Working Groups: Participation in developing specifications and guidelines, directly shaping Decentralized AI standards
- Events and Conferences: Regular summits, webinars, and educational activities
- Open Source Development: Open source and open governance project structure with modular, decentralized stack for the agentic era
Future Directions
The project aims to position participants at the forefront of creating the internet of AI Agents, providing exposure and faster adoption for AI products and tools. Major milestones include ongoing events at the MIT Media Lab and continued development of the foundational protocols that will enable the next generation of distributed artificial intelligence.
Contact and Participation
- Project Website: nanda.media.mit.edu
- Research Group: MIT Media Lab, Camera Culture Group
- Principal Investigator: Prof. Ramesh Raskar
- Researchers: Pradyumna Chari, Ayush Chopra
NANDA represents MIT's commitment to pioneering the infrastructure that will define how AI agents interact, collaborate, and scale in the decentralized future of artificial intelligence.