google/agentmesh

SAM is a peer-to-peer networking framework that enables autonomous agents to communicate and coordinate securely across distributed systems.

The project addresses the challenge of building trustworthy agent-to-agent communication in decentralized environments. SAM provides a mesh network architecture where agents can discover, authenticate, and exchange messages with one another without relying on centralized intermediaries. The framework emphasizes sovereignty, allowing each agent to maintain control over its own identity and communication policies while participating in a larger network of collaborating agents.

Developers building multi-agent systems that require decentralized coordination should consider SAM when agents need to operate across network boundaries without a central authority managing connections. The tool suits scenarios where agent autonomy and network resilience matter more than centralized control, such as distributed AI systems, peer-to-peer agent networks, or security-sensitive deployments where no single point of failure is acceptable. The project integrates with the Model Context Protocol, positioning it within the broader ecosystem of agent communication standards.

The project shows active development with regular commits addressing core networking functionality and agent coordination features. Work spans implementation of mesh topology management, peer discovery mechanisms, and security protocols for agent authentication. The codebase reflects ongoing refinement of the peer-to-peer architecture and integration patterns with agent frameworks.