The Decentralized Universal (or Agent) Discovery Protocol (DUADP) represents a foundational infrastructure layer within the emerging agentic computing ecosystem. It is designed to enable secure, federated discovery and interaction between AI agents, services, and capabilities across organizational and network boundaries, without reliance on centralized registries.
At its core, DUADP functions analogously to the Domain Name System (DNS), but instead of resolving websites, it resolves agents, tools, and executable capabilities distributed across domains.
DUADP addresses a critical structural gap: as AI systems evolve into autonomous, distributed agents, there is no universal mechanism for discovering, verifying, and interacting with them across environments.
The protocol introduces a decentralized discovery layer that allows:
Agents to locate other agents dynamically
Organizations to publish capabilities across domains
Systems to interact without preconfigured integrations
Federated ecosystems to replace siloed agent platforms
This establishes DUADP as a network-level discovery protocol for the “Internet of AI Agents.”
DUADP is built on a set of pragmatic and extensible architectural principles that leverage existing internet infrastructure while introducing agent-specific discovery capabilities.
DUADP uses domain ownership as the root of trust and discovery. Organizations publish agent endpoints associated with their domains, allowing other systems to query and resolve available agents.
This approach:
avoids centralized registries
aligns with existing internet governance models
enables verifiable ownership of agent endpoints
Instead of a single authority, DUADP relies on distributed discovery nodes that:
register agent metadata
respond to discovery queries
share capability information across networks
This creates a federated discovery fabric, ensuring resilience, scalability, and organizational autonomy.
Organizations expose agents through standardized endpoints that describe:
available capabilities (“skills”)
input/output schemas
execution interfaces
access requirements
This transforms agents into discoverable, callable services across domains.
The protocol uses simple HTTP-based APIs to:
register agents
query available capabilities
retrieve metadata
initiate interactions
This ensures compatibility with:
cloud-native systems
Kubernetes-based infrastructures
enterprise service architectures
Early implementations demonstrate SDK support in:
TypeScript
Python
Go
These SDKs enable developers to integrate discovery capabilities directly into agent runtimes and orchestration systems.
Within a production architecture, DUADP enables several critical functions:
Agents can locate other agents at runtime based on:
capability requirements
domain trust relationships
policy constraints
Organizations can expose services externally while maintaining full control over access and governance.
Rather than hardcoding integrations, systems can query:
“Which agent can perform this task?”
“What capabilities exist within this domain?”
DUADP enables multi-agent workflows, where agents dynamically assemble other agents into execution chains.
DUADP does not operate in isolation. It is part of a broader architectural stack:
DUADP discovers agents
OSSA defines their identity, capabilities, and permissions
Together, they provide discovery + governance
DUADP locates agents and services
Contract planes define how agents interact under enforceable agreements
This enables trusted execution across organizational boundaries
Within the Evidence Platform architecture, DUADP plays a critical role in enabling federated agent interaction across sovereign environments:
Each institution (law enforcement, defense, courts) operates independent Kubernetes clusters
Agents within those clusters remain data-sovereign
DUADP allows these agents to discover external capabilities without exposing internal systems
Example use cases:
A defense-side AI agent discovers a publicly available forensic analysis agent
A prosecutorial system discovers compliance auditing agents
Oversight systems dynamically locate monitoring tools across jurisdictions
This aligns with the platform’s principle of:
centralized evidence integrity (Origin of Truth)
decentralized computation and intelligence
DUADP’s decentralized nature introduces both advantages and requirements:
eliminates single points of failure
reduces centralized attack surfaces
supports domain-based trust models
agent identity verification (e.g., via OSSA / DID)
signed metadata and endpoints
policy-based access control
audit logging of discovery and invocation events
When combined with OSSA, DUADP supports secure, verifiable agent discovery rather than anonymous service exposure.
DUADP represents a shift from static service integration to dynamic, discoverable agent ecosystems.
If broadly adopted, it enables:
decentralized AI marketplaces
autonomous multi-agent collaboration
cross-organizational workflow orchestration
composable “agent-native” infrastructure
This positions DUADP as a potential foundational protocol for agentic computing, similar in importance to DNS for the traditional web.
The Decentralized Universal (Agent) Discovery Protocol establishes a federated discovery layer for AI agents, enabling systems to locate, evaluate, and interact with capabilities across domains without centralized control.
It transforms AI agents from isolated components into network-addressable, discoverable services, forming the backbone of a distributed, interoperable agent ecosystem.
DUADP Project Overview — https://duadp.org
Kubernetes — foundational infrastructure for agent deployment
Domain Name System — conceptual analogue for discovery
W3C Decentralized Identifiers (DID) Core — identity layer commonly paired with agent systems
National Institute of Standards and Technology — guidance on distributed systems and AI governance