Currently, integrating an AI agent with the Agent Mesh requires running the sam-node Go binary as a separate background daemon or Kubernetes sidecar. While this successfully isolates the libp2p transport and Biscuit authorization logic, it introduces significant developer friction. It forces developers to manage localhost IPC/ports, breaks lightweight serverless deployments (e.g., Cloud Run, Modal, AWS Lambda), and fragments OpenTelemetry tracing between the agent application and the network layer.
Since libp2p is designed to be embeddable and already has mature implementations in both JavaScript (js-libp2p) and Rust (rust-libp2p), we request native, in-memory SAM SDKs for these languages.
- TypeScript/Node: Combine js-libp2p with a WebAssembly (Wasm) compiled Biscuit Datalog engine so Node/Edge agents can evaluate mesh policies in-memory.
- Rust (and Python via PyO3): Leverage rust-libp2p alongside the native Rust biscuit-auth crate for embedded, high-performance edge agents.
This SDK-first approach would allow developers to connect to the mesh and expose Model Context Protocol (MCP) tools within a single process space (e.g., const mesh = new AgentMesh(...)), completely eliminating the infrastructure overhead of sidecars.
Idially it should integrate seamless with agentic platforms https://northflank.com/blog/top-fly-io-sprites-alternatives-for-secure-ai-code-execution-and-sandboxed-environments and allow platform teams that use kubernetes or other self managed platform to integrate easily.
Currently, integrating an AI agent with the Agent Mesh requires running the sam-node Go binary as a separate background daemon or Kubernetes sidecar. While this successfully isolates the libp2p transport and Biscuit authorization logic, it introduces significant developer friction. It forces developers to manage localhost IPC/ports, breaks lightweight serverless deployments (e.g., Cloud Run, Modal, AWS Lambda), and fragments OpenTelemetry tracing between the agent application and the network layer.
Since libp2p is designed to be embeddable and already has mature implementations in both JavaScript (js-libp2p) and Rust (rust-libp2p), we request native, in-memory SAM SDKs for these languages.
This SDK-first approach would allow developers to connect to the mesh and expose Model Context Protocol (MCP) tools within a single process space (e.g., const mesh = new AgentMesh(...)), completely eliminating the infrastructure overhead of sidecars.
Idially it should integrate seamless with agentic platforms https://northflank.com/blog/top-fly-io-sprites-alternatives-for-secure-ai-code-execution-and-sandboxed-environments and allow platform teams that use kubernetes or other self managed platform to integrate easily.