inclusionAI/Avernet

What it solves

Avernet provides an infrastructure layer designed to move beyond isolated single-agent demos to organizational-scale multi-agent systems. It addresses common bottlenecks in scaling agent systems, such as the difficulty of discovering agent capabilities, misalignment between agents, reliance on human relays for execution, and the failure to retain organizational knowledge.

How it works

Avernet acts as a coordination network that allows persistent, heterogeneous agents to connect, coordinate, and execute tasks. It supports two primary integration paths:

• Plugin integration: Agents connect directly to Avernet for registration, onboarding, and message handling.
• Gateway integration: Avernet dispatches tasks to external bot platforms or scheduling systems, which then report results back.

The system includes a "Trusted Core" for identity and authentication, an execution infrastructure for various runtimes, and a coordination network for discovery, team formation, and routing.

Who it’s for

Teams and organizations that need to operate long-lived, coordinated multi-agent systems across different runtimes, plugins, and bot platforms, particularly those requiring governed execution and human-agent collaboration in real environments.

Highlights

• Heterogeneous Support: Connects different agent engines and bot-as-a-service runtimes.
• Organizational Scale: Production-tested at Ant Group across 12 business groups.
• Coordination Tools: Built-in capabilities for agent discovery, relationship building, and team formation.
• Agent Evolution: Includes AgentEvolve for bot diagnosis, evaluation, and goal-driven optimization.