google

sam

Zero-trust P2P mesh for agents, published by Google: libp2p nodes with portable crypto identities self-discover, and an MCP sidecar lets an agent invoke tools on any node — cloud, local or edge.

643 95 Go Apache-2.0updated today

agents gateway security

View on GitHubDispute this mapping →

Curator's take

The bet here is transport, not agents. Portable cryptographic identity plus libp2p routing means an agent's tools stay reachable when it moves between laptop, cluster and edge, and every node, connection and packet is authenticated instead of trusted for being inside the VPC — plus a public testnet you can join in minutes and a `sam-mesh` skill that puts an agent on the mesh in one line. Costs: you operate a control plane, routers and per-host nodes, and Go/Kubernetes fluency is assumed. Read the disclaimer too — published by Google, but not an officially supported Google product and outside its vulnerability reward program. If your agents all live in one cluster and call a handful of MCP servers, a gateway is far less machinery.

Mapped by ShipWithAI editors · links verified

README.md

SAM: Sovereign Agent Mesh

SAM is a smart network built for autonomous AI agents:

Zero Config: Nodes discover each other and build the P2P network automatically.
Zero Trust: Every connection, node, and packet is strictly authenticated.
Agentic Network: Formed by lightweight nodes (sam-node) that provide self-healing, P2P connectivity, allowing autonomous agents to plug in, communicate, and invoke tools dynamically.
Portability: Cryptographic identities are environment-agnostic, allowing seamless node mobility across cloud, local, and edge environments.

Getting started is a one-liner (see the Quick Start Guide): install, add the skill, and your agent is on the mesh.

Demo: installing SAM, adding the sam-mesh skill, and an agent discovering and calling tools across the mesh

Advanced demo: an agent fans a batch of work across a warm pool of reviewer agents on the mesh

Full walkthrough: Warm Agent Pool use case.

Architecture Components

sam-control-plane: The registry control plane for node identity registration, authorization policies, and router coordinating.
sam-router: The libp2p bootstrap nodes and relays providing data-plane connectivity and forwarding.
sam-node: The local node clients providing mesh transport integration and MCP sidecar routing.

Documentation

Start exploring the Sovereign Agent Mesh:

For Users & Operators

Get a node running on the public testnet (bananas.sam-mesh.dev) in minutes:

🚀 User Quick Start Guide: Connect and run a SAM node using binaries or Docker, and query the local MCP server.
🤖 Agent Integration Guides: Connect Google Gemini, Claude, and other AI agents to your SAM node to dynamically discover and call tools across the mesh.
📡 Testnet Validation Tutorial: Real-time verification, remote tool invocation, and HTTP stream proxies.
🎛️ Production Kubernetes Deployment: Run your own hub (control plane, router, console) via plain manifests, or via the sam-mesh Helm chart for local/self-hosted setups.

For Developers & Contributors

Compile from source, run local clusters, or execute tests:

🛠️ Developer Guide: Prereqs, compilation, local control plane setup, and Kubernetes Kind deployment.
🧪 Testing Guide: Go tests, E2E BATS, and containerized mesh execution.

License

See LICENSE.

Disclaimer

This is not an officially supported Google product. This project is not eligible for the Google Open Source Software Vulnerability Rewards Program.

Continue your stack

What teams reach for next — and why each earns a place beside sam. Ranked by curator confidence.

Pairs well with

agentfield

Open-source control plane that runs AI agents as microservices: write plain Python/Go/TS functions, get REST endpoints with routing, queues, retries, memory and tracing — one laptop to 10k agents.

Why it fitsagentfield turns agent functions into routed microservices on one control plane inside your infra; SAM is the zero-trust wire between such planes once the agents and their tools span cloud, local and edge hosts.

numbat

Perplexity's endpoint sensor for AI agents: hooks, OTLP logs and on-disk session artifacts normalized into one CEL-matched event model, with opt-in pre-action blocking and offline forensics.

Why it fitsDefense in depth for agent infrastructure: numbat watches what an agent actually did on the endpoint, SAM authenticates who it is on the wire and what it may reach.

Alternative to See all + compare →

mcp-context-forge

IBM's registry + proxy federating MCP, A2A and REST/gRPC into one governed endpoint: discovery, auth, rate limits, guardrails, OTel tracing, 40+ plugins — scales to K8s with Redis federation.

Why it fitsTwo answers to 'how does an agent reach many tool servers': IBM's forge federates MCP/A2A/REST behind one governed central endpoint; SAM removes the center, giving every node an authenticated identity on a self-healing P2P mesh.