sirkirby/unifi-mcp

What it solves

UniFi MCP allows AI assistants and automation tools to interact with Ubiquiti UniFi controllers. It bridges the gap between natural language interfaces and complex network infrastructure management, enabling users to query, audit, and modify their UniFi Network, Protect, and Access deployments using AI agents.

How it works

The project implements the Model Context Protocol (MCP), exposing UniFi controller functionality as a set of queryable and composable tools. It consists of several specialized servers (Network, Protect, and Access) that connect to local controllers via admin accounts.

For remote access, it uses a "Cloud Relay" system: a Cloudflare-hosted Worker gateway acts as an authenticated edge endpoint, which communicates via a WebSocket to a local Relay sidecar that forwards requests to the local MCP servers. It also includes an independent REST and GraphQL API server for non-MCP consumers.

Who it’s for

It is designed for network administrators and home lab enthusiasts who use UniFi hardware and want to manage their infrastructure through AI-powered agents (like Claude Code, Codex, or OpenClaw) rather than manual dashboard navigation.

Highlights

• Comprehensive Toolset: Provides hundreds of tools across Network (192), Protect (61), and Access (36) domains.
• Agent Skills: Includes pre-defined skills for complex tasks like network health checks, firewall auditing with 100-point scoring, and security digests.
• Safety First: Implements a "preview-then-confirm" flow for all mutations and automatically redacts sensitive fields like Wi-Fi passphrases and VPN keys from agent context.
• Flexible Deployment: Supports direct local execution, plugin-based installation for AI IDEs, and a cloud-relay architecture for remote management.