Links

Sign up for updates: Sign-up

Email: dec-ai@media.mit.edu

Events Calendar: https://lu.ma/nanda

Join the conversation: LinkedIn

NANDA RoadMap

Learn More

How to Support

The program is financially supported via the AGNI research projects.

AI Agents and Networked Intelligence | MIT Media Lab

Principal Investigator: Professor Ramesh Raskar, Camera Culture Group

Overview

AI is transitioning from systems that generate content to systems that act. AI agents are emerging as entities that can plan, reason, communicate, and execute tasks on behalf of individuals, organizations, and computational systems. As these agents begin to interact with one another, a new computational layer is forming: a decentralized Internet of AI Agents, or the Agentic Web. The AGNI research projects, at the MIT Media Lab, will study and build this emerging layer.

The AGNI research projects bring together researchers, students, and member organizations to investigate how AI agents are designed, deployed, discovered, trusted, coordinated, and evaluated across open and heterogeneous environments — combining fundamental research, system-building, and hands-on experimentation in shared agent environments.

Research Themes

The AGNI research projects are organized around two Research Themes. Each Research Theme is a self-contained research engagement — its own statement of work, timeline, and sponsor group.

Research Theme 1 — Foundations of the Agentic Web

Technical foundations for independent AI agents to operate across organizational and technical boundaries: agent identity, naming, and discovery; interoperable agent-to-agent communication; authentication, authorization, and secure delegation; provenance, trust signals, and policy-aware privacy; and deployment across cloud, edge, device, sensor, and embedded environments.

Core question: how can independent AI agents reliably discover, authenticate, communicate, and collaborate across decentralized environments?

Research Theme 2 — Agent Behavior, Markets, and Society

Emergent behavior when many AI agents interact across shared environments: multi-agent coordination, negotiation, and orchestration; population-scale dynamics, network effects, and reputation; agent-mediated commerce, pricing, and incentive design; and governance of systemic risks (collusion, manipulation, spam, accountability gaps).

Core question: what new forms of coordination, intelligence, and economic activity emerge when AI agents operate, interact, and transact at scale?

How Participation Works

The AGNI research projects run on the same mechanism as any Media Lab multi-sponsor FRP — they're simply organized under a shared name across two thematic Research Themes rather than presented as a single one-off project. Each sponsor on a given Research Theme will be on the same statement of work, the same period of performance, and the same terms.

In effect, sponsors joining a Research Theme form a class — a cohort that commits and comes online together. A Research Theme's class convenes once a sufficient group of sponsors is aligned and ready to commit; at that point, MIT executes the contract for the whole class at once. This is why timing matters: expressing interest early increases the chance of shaping which class you join and when it convenes.

Two classes are open now — one per Research Theme — and a sponsor may join either or both.

Prerequisite: participation requires active MIT Media Lab Consortium membership. Prospective sponsors who are not yet Consortium members should raise that in parallel with expressing interest in the AGNI projects.

Activities

• Research & Publications — new architectures, protocols, models, simulations, and prototypes for decentralized AI and agentic systems. Outputs may include research papers, technical reports, white papers, open-source systems, benchmarks, and reference implementations.
• AI Studio Course — students design, build, and deploy AI agents in shared environments, covering design, deployment, evaluation, and coordination; may culminate in a demo session where systems are tested in live or simulated settings.
• Agent Sandbox — an experimental environment where heterogeneous agents interact in controlled settings, supporting experiments in discovery, communication, interoperability, coordination, market behavior, and evaluation.
• Workshops & Research Sessions — technical workshops, research briefings, working sessions, roundtables, and periodic progress reviews.
• Shared Resources — sponsors receive access to the shared outputs of the Research Theme(s) they support: reports, prototypes, demonstrations, research briefings, and experimental resources.

Participation

AGNI sponsorship works per Research Theme..

Contribution: $50,000 per Research Theme.

• Support one Research Theme — $50,000
• Support both Research Themes — $100,000 (each theme's own scope, deliverables, and budget)

Sponsors of a Research Theme receive, for that theme:

• Participation in theme-specific research briefings, workshops, and review sessions
• Access to theme-specific reports, white papers, prototypes, demonstrations, and final presentations
• Opportunity to suggest use cases, research questions, or problem domains for the research team's consideration
• Opportunity to issue a white paper in collaboration with the AGNI team
• Visibility into emerging student and researcher work related to that theme

Sponsors supporting both Research Themes receive the combined benefits and deliverables of both.

Expected Outcomes

• Architectures for decentralized AI and agent-based systems
• Prototypes of interoperable agent ecosystems
• Experimental sandbox environments for agent interaction
• Models of multi-agent coordination and population-scale behavior
• Early models for agent-based interaction, exchange, pricing, and reputation
• White papers on the Agentic Web, agentic commerce, and agent ecosystem governance
• A new generation of students trained in designing, deploying, and evaluating agentic systems

Interested?

Contact the MIT Media Lab External Relations team to discuss which Research Theme fits your organization's interests and where the current class stands.