Teams that run AI agents across clouds, data centers, laptops and edge devices face a common security headache. To let agents call each other’s tools they must expose internal scripts, LLM endpoints or private APIs to the public internet. This creates a large attack surface and violates compliance rules in finance, health, government and industrial settings. The alternative is to build a private VPN‑like overlay that works only for agent‑to‑agent traffic, but traditional solutions need manual IP management, complex firewall rules and constant re‑configuration when nodes move or NAT changes.
SAM (Sovereign Agent Mesh) solves this problem with a zero‑config, zero‑trust peer‑to‑peer overlay built specifically for sharing Model Context Protocol tools. Three lightweight binaries handle identity, routing and node logic. A control plane verifies OIDC JWTs from your existing auth provider, converts the claims into Biscuit capability tokens and distributes policy. Each node then authorizes every request locally, using a strict default‑deny rule set that requires an explicit granted_service fact for every call. No service is reachable until a policy says so, and decisions are made offline without calling back to the control plane.
Deployment is straightforward: pull the Go binaries or Docker images, run the control plane (self‑hosted for full data sovereignty), then start routers and nodes with simple join and run commands. The mesh automatically discovers peers, heals broken links and survives NAT traversal using libp2p on UDP 5001 and TCP 5002. The local MCP API listens on port 8080, so existing agent code needs only a small configuration change to point to localhost.
For regulated organizations this means agents can securely share databases, inference endpoints or custom tools across cloud, on‑prem and edge without ever opening a port to the internet. Security teams retain full auditability because every request is cryptographically signed and can be inspected locally. Engineering teams gain a production‑ready, Apache‑2.0 licensed solution that scales from a few laptops to thousands of edge nodes.
#AI #Product #Security #DevOps #Cloud #EdgeComputing