Runtime Implementation
Provides the routing and publishing infrastructure for cross-network AI orchestration
Agents self-organize into coordination loops based on capability, not configuration. No hardcoded workflows—just voluntary contribution.
Loops spawn sub-loops infinitely. Each layer aggregates intelligence from the layer below. Collective intelligence emerges naturally.
Not just routing messages—creating new insights through multi-agent collaboration. The whole becomes greater than the sum of its parts.
Agents across different networks, organizations, and trust boundaries coordinate seamlessly through policy-driven routing.
No central authority decides. Agents vote, merge, and reach consensus through configurable aggregation strategies.
Every decision, every contribution, every loop execution is logged. Complete transparency and accountability at every level.
Manage your realm hierarchy, configure policies, and monitor agent activity through the Nexus Console interface.
Visualize and manage your nested realm structure with expandable tree navigation
Configure routing policies, inheritance rules, and access controls per realm
Track SDK clients, connection status, and agent activity in real-time
SDK clients connect to Nexus and join specific realms based on their role (service, event, loop participant). Each realm acts as a security boundary with its own policies.
Nexus routes service calls, publishes events, and broadcasts loop invitations across the network. Policy engines determine which agents can communicate based on capabilities and trust boundaries.
When a loop is initiated, Nexus broadcasts to all capable agents. Agents self-select to participate, execute in parallel, and Nexus aggregates results using the configured strategy (vote, merge, consensus).
Loops can spawn sub-loops, creating recursive coordination patterns. Each level aggregates intelligence from below, enabling emergent collective intelligence that scales infinitely.
The Nexus MVP demonstrates agents routing through realms in real-time. Watch as the Ping-Pong test agents coordinate through the runtime.
Test agents send ping and pong messages back and forth, demonstrating real-time routing through Nexus realms with full activity monitoring.
Track connections, events, and errors in real-time. Every agent interaction is logged with timestamps and connection status.
Realms are the foundational building blocks of InterRealm. Each realm acts as a security boundary and routing domain, defining what agents have access to and how they coordinate.
Realms form a parent-child hierarchy. Each realm can contain AI agents, logical resources, or other nested realms. Routing flows through this hierarchy based on policies.
Members (SDK clients) connect to realms and gain access based on realm policies. No explicit dependency injection—agents discover and coordinate dynamically.
Realms are defined using the InterRealm protocol schema. Each realm specifies its type (Root, Service, Organization), parent relationship, and member access rules. This schema enables:
Nested realms create organizational structure and routing paths
SDK clients connect and agents coordinate within realm boundaries
Define what agents can access and how they interact across realms
Members are connection points for external systems using the InterRealm SDK. They define how agents, services, and logical resources participate in the mesh with delegated security and dynamic routing.
Members act as slots for external systems to connect via the InterRealm SDK. Each member has a type (Consumer, Provider, or Hybrid) and optional contract specification.
Members inherit security context from their realm. Routing and access control are defined at the realm level, not hardcoded in agent logic.
This is where the magic happens. Agents declare what they can do (capabilities), not what they need (dependencies). Nexus handles routing, discovery, and coordination dynamically:
Agents find each other through capability matching, not explicit configuration
Coordination spans networks and organizations through realm-based routing
Add new agents without updating existing ones—they discover each other automatically
After creating a member, Nexus generates a unique API key for SDK authentication. This key is shown only once and must be stored securely by the client application.
API keys are generated once and never stored in plain text. After creation, the key is displayed only once—if lost, a new member must be created.
Use the API key in your SDK client configuration to authenticate and connect to the Nexus gateway. The SDK handles all authentication and connection management.
API keys are the only credential needed to connect to InterRealm. Treat them like passwords and follow these security guidelines:
Use environment variables or secret management systems—never commit keys to source control
Create new members periodically and deactivate old ones to minimize exposure risk
Create separate members for different services to isolate access and simplify revocation
Track connection activity in the Nexus Console to detect unauthorized access attempts
Explore the InterRealm Stack and start building distributed AI systems