The NANDA Index is the agent registry at the center of Project NANDA (Networked AI Agents in Decentralized Architecture), a research initiative that originated at MIT. The project targets four pieces of infrastructure it sees as choke points for AI agents on the open web: naming (the role DNS plays for websites), certificate authorities, orchestration, and attestation.
The index lets agents register, be discovered, and establish verifiable identity. Its design, along with the AgentFacts schema that describes an agent and a verified agent discovery process, is set out in the paper "Beyond DNS: Unlocking the Internet of AI Agents". Related papers cover an adaptive resolver for agent name resolution, zero trust agentic access for enterprise deployments, and a survey comparing NANDA with MCP, A2A, and Microsoft Entra as registry approaches.
The current phase of the roadmap covers index infrastructure, protocol bridges between A2A, MCP, and HTTPS, and SDKs for onboarding agents. Later phases plan mechanisms for agents to price and exchange knowledge, edge AI, economic protocols for agent services, and privacy-preserving co-learning across agent populations.
Features
- Agent registry: agents register in the NANDA Index and are discovered through it
- AgentFacts schema: a structured description of an agent used for discovery and verification
- Verified discovery: agents establish verifiable identity when they are found
- Adaptive resolver: a microservice architecture for resolving agent names in distributed environments
- Protocol bridges: planned bridges between A2A, MCP, and HTTPS
- Agent onboarding SDKs: SDKs for deploying and registering agents
- Research publications: papers on registry design, DNS limitations for agents, and enterprise zero trust access
