LabTetherArchitecture
How it works
LabTether is a hub-and-spoke architecture. The hub runs in Docker, agents run on your nodes, connectors bridge to your infrastructure platforms, and the AI gateway exposes it all via MCP.
The Hub
The hub is the central control plane. It runs as a Docker Compose stack (API server, web console, worker, Postgres) on any Linux host. All agents and connectors report to it. The web console provides real-time dashboards, alert management, remote terminal access, and update orchestration.
Agents
Lightweight Go binaries that run on every node you want to manage. They report telemetry (CPU, memory, disk, network), provide terminal/desktop access, execute remote commands, and handle file operations. Available for Linux, macOS, and Windows with native wrappers for each platform.
Connectors
Connectors bridge LabTether to your infrastructure platforms via their native APIs. Proxmox VE, TrueNAS, Docker, Portainer, Home Assistant, and Proxmox Backup Server are supported. Each connector normalizes platform-specific operations into a unified model — so the same dashboard and AI gateway work across everything.
AI Agent Gateway
The MCP (Model Context Protocol) gateway exposes your entire infrastructure to AI clients. Any MCP-compatible tool — Claude, Cursor, OpenClaw — can discover your nodes, query metrics, run diagnostics, and execute actions through natural language. The gateway enforces the same RBAC and audit logging as the web console.
Security Model
All agent-to-hub communication uses mutual TLS. The hub supports OIDC SSO and role-based access control. Every action — whether from the console, CLI, mobile app, or AI agent — is logged in a tamper-evident audit trail. The MCP gateway inherits the same permission model, so your AI agent can only do what your user role allows.