Introducing SonicWall MCP: AI-Powered Network Management

Picture this: you need a new access rule: WAN to LAN, allow SSH, from one specific host. It takes ten seconds to tell someone what you need. Why should it take significantly longer to make that change? SonicWall MCP closes that gap, letting you describe the configuration in plain language and get it done in a fraction of the time.

We're releasing “SonicWall MCP” as an open-source server that lets AI assistants like Claude, Gemini, ChatGPT, and others talk directly to SonicWall's products: Firewalls (via Network Security Manager (NSM) and SonicOS), Cloud Secure Edge (CSE), Capture Client, and MySonicWall (MSW), all in plain English.

The Problem

Cybersecurity products are often designed with security professionals in mind. These are the folks who already understand the terminology, workflows, and menu structure across multiple consoles. While the UI works right out-of-the-box for them, it falls short for anyone who needs a fast answer or a fast change: an admin configuring policy under time pressure, a manager who wants a plain language status update, or a developer who just wants to script one thing without reading four hundred pages of API docs.

How SonicWall MCP Solves the Problem

SonicWall MCP connects an AI assistant to the entire API surface using the Model Context Protocol (MCP). MCP is an open standard for allowing AI tools to use external systems safely. It works within the boundaries of what each product's API exposes; it doesn't invent capabilities the API doesn't have. Within those boundaries, three things fall out naturally:

• Configure with intent. Describe what you want, for example a new access rule, an address object, or a firmware upgrade, and the assistant translates that into the right sequence of API calls, eliminating the need to navigate it by hand.
• Fetch whatever the API can tell you. Devices, licenses, pending changes, policies, and endpoint stats: any data the underlying product exposes is a plain language question.
• Perform the operations the API supports. Anything from staging a config change to running a firmware upgrade: if the API can do it, it's reachable through the assistant.

Instead of manually building a separate tool for each of those thousands of operations, which would be unmanageable, the server indexes the products' own API specifications and exposes a small number of smart lookup tools that can search, describe, and safely call any operation on demand. A handful of the most common actions, such as logging in, listing devices, and reviewing pending changes, result in a more polished, purpose-built path on top of that.

One conversation for every product

SonicWall MCP gives administrators a faster, more intuitive way to interact with their entire SonicWall environment. Built into Unified Management, MCP provides a natural-language interface for monitoring, troubleshooting, configuring, and managing security and networking infrastructure.

Instead of navigating through menus or looking for a specific setting across different solutions, administrators can simply ask for what they need in plain English. Check out a firewall rule, investigate an endpoint’s status, review a device policy, or make a configuration change, all from the same conversation.

Unified Management brings the environment together. MCP makes it conversational.

Why does this matter beyond convenience?

For a developer, this means scripting SonicWall workflows without wrestling with raw API calls. For an IT team, it means describing an intent, "open SSH from this host"; instead of building it by hand, click-by-click. For a decision maker, it means the gap between "we need this change" and "it's done" gets meaningfully shorter, without adding a new source of risk, because the safety rails are built in, not bolted on.

Turn intent into action in seconds

SonicWall MCP is open source under the MIT license: free to use, free to fork, free to build on, and ready to connect to Claude Code, Claude Desktop, or any MCP-compatible assistant. Setup takes one command and one API key: no infrastructure to stand up, nothing to provision beyond what you already have. Try it here.

Grab it, point it at a test tenant, and ask it something. That's the whole pitch: stop clicking through consoles and start conversing.