Project NANDA: Architecting the Internet of AI Agents

Project NANDA: Building a Quilt for the Internet of AI Agents How can trillions of AI agents collaborate across a organizational silos: communicate seamlessly, navigate autonomously, socialize, learn, earn and transact on our behalf. Project NANDA is building a quilt for a new kind of internet i.e. The Internet of AI Agents. NANDA, created at MIT, provides the index of federated agent registries, communication interface, and open-source tooling needed to enable this system in a decentralized, protocol-neutral way. Overview Project Nanda roadmap includes many research and implementation problems: agent discovery, Packet Switched Intelligence, trust, privateML, knowledge pricing, population AI and more. In the short term, NANDA defines a minimal set of abstractions to: Identify agents across the open web Describe agents with structured facts and capabilities Enable communication across protocol boundaries (A2A, MCP, HTTPS) Deploy agents in managed or self-hosted environments Interact with agents through open clients or UX layers (e.g. chat, workflows) The architecture is intentionally modular and protocol-agnostic. NANDA does not define a new transport layer. Instead, it stitches together existing and emerging protocols through open registries and SDKs. Protocol Interoperability: MCP, A2A, HTTPS Current Reference Implementations 1. NANDA Index : https://index.projectnanda.org 2. AgentFacts: https://list39.org 3. Agent Picker SDK: https://github.com/projnanda/nanda-sdk 4. Chat Interface: https://chat39.org LinkedIn Channel: https://www.linkedin.com/company/projectnanda/ Events Calendar: https://Lu.Ma/nanda Youtube Channel: https://www.youtube.com/@ProjectNANDA GitHub repos: https://github.com/projnanda Onboarding steps for NANDA writing group: https://tinyurl.com/nandawritingonboarding Connect with & Sign up for newsletter: https://forms.gle/bnVRV2zr8M5Kotfr7 Homepage: https://projectnanda.org MIT research group: https://nanda.mit.edu