WatchGuard MCP

WatchGuard + AI - for ChatGPT, Claude, GitHub Copilot, Microsoft 365 Copilot, Gemini, and any agent that speaks MCP

Unofficial. Community-built Claude Code Skill and MCP server for the WatchGuard API. Not affiliated with, endorsed by, or sponsored by WatchGuard Technologies, Inc.

_{▶ Watch the 30-second demo - demo data is simulated; every command shown exists in the real CLI.}

WatchGuard Cloud's Firebox, FireCloud, Endpoint Security, ThreatSync, NDR, AuthPoint and account platform APIs in one CLI and MCP server, with cross-customer reports for Service Providers. Works with the AI you already use - ChatGPT (Plus/Pro+, paused until this connector is reprinted; see mcp-install.md), Claude Desktop, Codex, Claude Code, Claude Cowork, and GitHub Copilot - plus Microsoft 365 Copilot / Copilot Studio and Google Gemini via the remote path. Free, open source, runs on your laptop. Built for MSP owners. No code required.

Works with your agent

Security note (until this connector is reprinted). Do not run HTTP mode (--transport http), not even on localhost. This build checks a token, but it predates the fix release's request size and time limits. The MCP tools can also make changes with your credentials. Use the default stdio transport. If you set up a tunnel for a remote agent, shut it down. A fix is on the way.

The six agents MSP owners actually use (self-serve, works today):

Your AI agent	How to install the WatchGuard skill
Claude Desktop	Run installer, then Settings > Extensions to register watchguard-mcp (no JSON editing).
ChatGPT (paid plans)	Paused until this connector is reprinted: ChatGPT needs its HTTP mode, and in this build that mode checks a token but has no request size or time limits yet. See the security note in mcp-install.md.
Codex CLI	Paste the install prompt below.
Claude Code	Paste the install prompt below.
Claude Cowork	Paste the install prompt below.
GitHub Copilot (VS Code)	Run installer, add watchguard-mcp to mcp.json under the servers key, then pick Agent mode.

ChatGPT needs a remote HTTPS server, which means this connector's HTTP mode, and that is paused until the connector is reprinted: in this build that mode checks a token but has no request size or time limits yet. See the security note in mcp-install.md.

Also for the Microsoft and Google stacks

Big install base, but an honest heads-up: these are the remote / enterprise path, not the local binary you just installed.

Agent	What it takes
Microsoft 365 Copilot / Copilot Studio	Paused until this connector is reprinted: Copilot Studio needs its HTTP mode, and in this build that mode checks a token but has no request size or time limits yet. See the security note in mcp-install.md.
Google Gemini	Gemini CLI is local - same as Claude Code. The Gemini app is remote and needs the same paused HTTP mode as ChatGPT. See mcp-install.md.

Skill-native agents (also covered): Hermes and OpenClaw read this skill's SKILL.md directly and speak MCP - see their install sections below. Also works with Cursor, Windsurf, Cline, Continue.dev, and Zed via MCP. Full per-tool wire-up: docs/which-agent.md.

Run more than one agent? Install across all 51+ supported agents in one command: npx skills add Servosity/msp-skills@latest (requires Node.js, then run the per-skill installer for the CLI/MCP binaries). See docs/which-agent.md.

Install in 60 seconds

Fastest for Claude Desktop - one-click .mcpb

Download WatchGuard MCP (.mcpb) - then open Claude Desktop > Settings > Extensions and select the file. One click, no JSON, no shell. (Browse every WatchGuard release on the releases page.)

Prefer the Claude Code plugin? Add the marketplace once, then install - works immediately, no directory listing required:

/plugin marketplace add Servosity/msp-skills
/plugin install watchguard@msp-skills

Path A - paste one prompt into your AI agent (recommended)

Copy this into Claude Code, Codex CLI, or Claude Cowork:

Install the WatchGuard Skill and MCP server from Servosity/msp-skills in this agent workspace. If this workspace uses a POSIX shell (macOS, Linux, WSL, or Bash), run bash <(curl -fsSL https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/watchguard/install.sh). If it uses Windows PowerShell, run iwr -useb https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/watchguard/install.ps1 | iex. Then authenticate per the README and run watchguard-cli --help to explore.

The same prompt works in any agent that can run shell.

Path B - run the installer yourself

Windows (PowerShell):

iwr -useb https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/watchguard/install.ps1 | iex

macOS / Linux:

bash <(curl -fsSL https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/watchguard/install.sh)

The installer drops both watchguard-cli (the CLI) and watchguard-mcp (the MCP server) into your user bin path. Claude Code, Codex, and Cowork discover the Skill via SKILL.md in this directory.

Verify:

watchguard-cli --version

Upgrade to the latest version

The installer always fetches the current release - re-run it to upgrade:

macOS / Linux:

bash <(curl -fsSL https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/watchguard/install.sh)

Windows (PowerShell):

iwr -useb https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/watchguard/install.ps1 | iex

Claude Desktop .mcpb users: download the latest .mcpb (top of this section) and re-select it in Settings > Extensions. Claude Code plugin users: /plugin update watchguard@msp-skills.

Add to Claude Desktop, GitHub Copilot, Gemini CLI, Microsoft 365 Copilot, or another MCP client

After the installer runs, see mcp-install.md and docs/which-agent.md for the per-agent wire-up - one section per agent, including the GitHub Copilot servers key and the remote Microsoft 365 Copilot / Copilot Studio path. Claude Desktop's Settings > Extensions panel is the simplest path; the MCP config block (for users who prefer editing JSON) is documented in mcp-install.md.

Install for Hermes

From the Hermes CLI:

hermes skills install servosity/msp-skills/skills/watchguard --force

Inside a Hermes chat session:

/skills install servosity/msp-skills/skills/watchguard --force

Hermes speaks MCP natively, so it can also use the watchguard-mcp server directly - same install path, same env vars.

Install for OpenClaw

Tell your OpenClaw agent (copy this):

Install the watchguard skill from https://github.com/servosity/msp-skills/tree/main/skills/watchguard. The skill defines how its required CLI (watchguard-cli) can be installed via the openclaw: frontmatter block.

OpenClaw isn't generally available yet; the frontmatter wiring is pre-shipped and will activate the moment OpenClaw launches.

Authenticate

Set the credentials the CLI needs (from your WatchGuard portal):

WATCHGUARD_ACCOUNT_ID=<value> WATCHGUARD_API_KEY=<value> WATCHGUARD_CLIENT_ID=<value> WATCHGUARD_CLIENT_SECRET=<value> watchguard-cli doctor

doctor checks config, paths and API reachability, and with the Access ID, password and API key set it requests a token and reads your account record, so a wrong password shows up as a Credentials: ERROR row (while a token minted earlier with the right password is still cached, doctor reuses it for up to an hour; run watchguard-cli auth logout first to recheck the password). It exits 0 even then, so scripts should add --fail-on error. A read command such as watchguard-cli accounts list-managed --agent is the end-to-end check.

What this skill does

Question your MSP keeps asking	Command
Which customer needs attention first across Firebox, endpoints, ThreatSync and licenses?	watchguard-cli accounts health --agent
Which endpoints are unprotected, out of date, isolated or waiting on a reboot, across every customer?	watchguard-cli endpoint-devices posture --agent
Which new ThreatSync incidents from the last 30 days need triage first, and which has nobody touched?	watchguard-cli threatsync-incidents queue --status new --duration 30D --agent
Which customers and computers are exposed to this CVE?	watchguard-cli endpoint-patches exposure --cve CVE-2024-38063 --agent
Which Fireboxes and access points went offline, came back or changed firmware since the last sync?	watchguard-cli firebox-devices changes --agent
Which Fireboxes run older firmware than the rest of the same model in my fleet?	watchguard-cli firebox-devices firmware-lag --agent
Which allocations don't match the Fireboxes that actually report in, and what expires this month?	watchguard-cli allocations reconcile --expiring-within 30d --agent
Did the Firebox deployment finish, or fail?	watchguard-cli firebox-transactions wait --device FB-12345 --max-wait 10m

accounts health, endpoint-devices posture and endpoint-patches exposure are built for a WatchGuard Service Provider account: they reach each customer with its own token. firebox-devices changes and firmware-lag read the local copy, so sync the device list first: watchguard-cli sync --resources firebox_devices --full (it includes your customers' Fireboxes by default). firebox-transactions wait exits 0 when the deployment completes, 8 when it fails, times out or is canceled, and 9 when it is still running at --max-wait.

Full command reference: guide.md. For the AI-agent operating contract (--agent, --dry-run, when to confirm before mutating), see AGENTS.md.

What makes this different

The only other MCP server for WatchGuard Cloud we found on GitHub as of October 2026, WYRE-AI/watchguard-cloud-mcp, covers the platform's account, activation and operator APIs and lists Firebox and Endpoint Security as out of scope for now. A question like "which of my 40 customers has an offline Firebox, an unprotected laptop and an untouched incident" still means switching into every account.

This skill covers 208 operations from 13 of WatchGuard's published API specifications: Firebox Management and Reports, FireCloud, Endpoint Security, ThreatSync, NDR assets and Smart Alerts, AuthPoint authentication, and the platform's accounts, operators, activations, license allocation and authorization APIs. Where WatchGuard requires it, the cross-customer commands mint a separate token for each managed account, so one Service Provider credential reaches every customer, and it syncs WatchGuard into a local SQLite mirror with full-text search. Compound commands like accounts health, allocations reconcile and firebox-devices changes join Fireboxes, endpoints, incidents, allocations and earlier snapshots of your fleet - work a stateless API wrapper can't do. WatchGuard Cloud stays where you configure Fireboxes and policies, and ThreatSync keeps doing its own correlation; this skill answers the cross-customer questions from the agent you already use.

The pain this closes

Firebox Management, Endpoint Security, ThreatSync, NDR and license allocation are separate WatchGuard APIs, each with its own paths, paging and status codes. Firebox and ThreatSync can answer across customers with a hierarchy or tenants flag, but WatchGuard Endpoint Security answers for a managed customer only with an access token minted for that customer: you look up the account's audience value and request a token with it (WatchGuard API docs, Authentication). WatchGuard's own blog celebrated Service Provider-level firmware upgrades because MSPs "no longer need to go down the line on each subscriber account, one at a time" (September 2022). And allocations drift from reality: a WatchGuard Community thread (May 2023, still asked about in September 2024) shows retired, unallocated Fireboxes inflating the inventory with a support case as the only way out.

• watchguard-cli accounts health - one ranked row per customer: offline Fireboxes, at-risk endpoints, open incidents by severity, expiring licenses.
• watchguard-cli endpoint-devices posture - every unprotected, outdated, isolated, unlicensed or reboot-pending computer, across every customer.
• watchguard-cli threatsync-incidents queue - open incidents ranked by severity, threat score and age, with untouched ones and recommended-but-not-taken actions called out.
• watchguard-cli allocations reconcile - allocated serials with no reporting Firebox, reporting Fireboxes with no allocation, and what expires soon.
• watchguard-cli firebox-transactions wait - a pass or fail exit code for a deployment instead of refreshing a status page.

See pain-point.md for the longer narrative.

Frequently asked questions

Does this work with ChatGPT?

Not with this build. ChatGPT (Plus, Pro, Team, Business, Enterprise, and Education plans; Free does not yet expose Developer Mode) connects only to remote MCP servers over HTTPS, which means running the WatchGuard MCP server in HTTP mode. In this build HTTP mode checks a token but has no request size or time limits yet, so do not run it or put it behind a tunnel until the connector is reprinted; the security note in mcp-install.md explains why. Until then, use Claude Desktop, Claude Code or Codex.

Does this work with Codex, Cursor, Windsurf, Cline, Copilot, or Gemini?

Yes - all of them speak MCP. Cross-tool install commands are in the matrix above and the deep-dive in docs/which-agent.md.

Do I need to know how to code?

No. The recommended install is to paste one sentence into Claude Code or Codex - your agent reads SKILL.md and does the install. The fallback is a one-line installer per OS (bash or PowerShell). Neither path requires writing code. You'll enter your WatchGuard credentials once.

Is my WatchGuard data safe?

The CLI, MCP server and SQLite mirror run on your machine, under your user account, and MSP Skills never bundles or transmits your WatchGuard credentials. What your agent reads does go to your AI provider: query results, and the secrets that credential reads return (BOVPN pre-shared keys, private keys, Firebox passwords, activation and license keys). Keep those commands human-in-the-loop as the safety tiers below describe.

What credentials do I need?

WatchGuard Cloud API access: as an Owner or Administrator, open Administration > Managed Access and click Enable API Access (a Service Provider does this from Subscriber view of its own account). You get a read-write Access ID, a read-only Access ID, the passwords you set for each, and an API key. Set WATCHGUARD_CLIENT_ID to an Access ID, WATCHGUARD_CLIENT_SECRET to its password, WATCHGUARD_API_KEY to the API key and WATCHGUARD_ACCOUNT_ID to your account ID, then run watchguard-cli doctor. For a reporting-only agent, use the read-only Access ID.

Does it work across all my customers, or one account at a time?

Across all of them, with a Service Provider account. accounts health, endpoint-devices posture and endpoint-patches exposure list the accounts you manage, mint a separate token for each one and report every customer in one result, naming any account they could not reach. threatsync-incidents queue reads incidents for all your managed accounts in one ThreatSync call. If any read fails, including a product a customer does not have (WatchGuard answers that like any other refusal), these commands exit 6 (5 when nothing could be read) and say which account and product; accounts health lists that account first with complete: false instead of showing zeros, and an account it could not reach at all appears in fetch_failures. Add --allow-partial-failure to exit 0 on a partial result; it does not override exit 5. To work on one customer, set WATCHGUARD_AUDIENCE to that customer's audience value (watchguard-cli audiences --account-id <customer-account-id> returns it) and WATCHGUARD_ACCOUNT_ID to that customer's account ID. That is a convenience, not a security boundary: WatchGuard's audience token also works on your own account.

My WatchGuard Cloud account is in Europe or Asia Pacific. Does it work?

Yes. Set WATCHGUARD_BASE_URL and WATCHGUARD_TOKEN_URL to your region's gateway: https://api.deu.cloud.watchguard.com for EMEA or https://api.jpn.cloud.watchguard.com for APAC, with /oauth/token on the end for the token URL. The default is the North America gateway, api.usa.cloud.watchguard.com. Managed Access in WatchGuard Cloud shows your account's API URL as https://api.<region>.cloud.watchguard.com/rest/; drop the /rest/ when you set WATCHGUARD_BASE_URL (the CLI adds it to every path).

Will this hit my WatchGuard API rate limits?

WatchGuard allows 500 requests per second and 200,000 requests per day per API key; a 429 means the day's quota is spent until 00:00 UTC. accounts health, endpoint-devices posture and endpoint-patches exposure fetch at most four accounts in parallel by default (--concurrency), the cross-customer commands stop at --max-pages, --rate-limit caps requests per second, GET responses are cached for 5 minutes, and sync keeps a local copy that search and analytics read without calling the API.

Can my AI isolate devices or change firewall policy?

Only if you let it. Write commands such as endpoint-devices isolate, endpoint-devices uninstall, firebox-deployments and the delete commands send as soon as they run; there is no confirmation prompt. Preview any of them with --dry-run first; through the MCP server there is no preview, watchguard_execute sends a write as soon as it is called. For a reporting agent, use the read-only Access ID, or set WATCHGUARD_MCP_READ_ONLY=1 so the MCP server drops its write tools and refuses every non-GET call through watchguard_execute.

Has it been tested against a live WatchGuard Cloud account?

Not yet. Every command was built from WatchGuard's published OpenAPI specifications and tested against a local mock of those APIs, not a real account. If you run it against your own WatchGuard Cloud account, tell us what worked or broke on the issue tracker or at a Build Session; a real MSP run is what earns the live-verified badge.

What does it cost?

Free. Apache-2.0 licensed. You pay only for whichever AI agent you use (Claude, ChatGPT, Codex, etc.), and that's billed by your AI provider, not by us.

Safety model

Tier	Examples	Recommended agent policy
Read	accounts health, endpoint-devices posture, threatsync-incidents queue, endpoint-patches exposure, allocations reconcile, firebox-devices changes / firmware-lag, firebox-transactions wait, sync, search, analytics, every list / get command except the credential reads in the last row	Allow
Endpoint and incident actions	endpoint-devices isolate / stop-isolation / scan / action / uninstall, endpoint-configurations assign, threatsync-actions perform, threatsync-incident-actions perform, threatsync-transactions, threatsync-incidents set-status, ndr-smart-alerts close, authpoint push	Human-in-the-loop, explicit confirmation
Firewall config and deployment	firebox-deployments, firebox-policies create / update / set-enabled, the Firebox and FireCloud exception creates and updates, firebox-exceptions enable-global / disable-global, firebox-templates subscribe, firebox-bovpn-p1-settings create / update, firebox-transactions update, endpoint-risks update-config, ndr-assets create / update, threatsync-comments create / update, allocations allocate / update-hardware / update-software, accounts create / update	Preview with --dry-run, then a human-approved write
Destructive	accounts delete (it targets --account-id, which defaults to your own WATCHGUARD_ACCOUNT_ID), operators delete, allocations deallocate, firebox-transactions cancel, ndr-assets delete, firebox-policies delete, threatsync-comments delete, and every other delete command	Human-in-the-loop only, explicit confirmation
Credential and admin	firebox-bovpn-tunnels list / get and writes (pre-shared keys), firebox-certificates list / get / create / install (private keys and passwords), firebox-networks, firebox-users and firebox-content-filtering list / get (PPPoE, Dynamic DNS, Wi-Fi and override passwords), activations list-recent / status / activate (activation keys), allocations list-owned / list-assigned / summary (serial numbers and license keys), export activations (activation keys), import, endpoint-installers (enrollment download links), operators create / update, audiences, the authpoint password, OTP, QR code, forgot-token, check-policy and transaction-status calls, auth login, auth set-token	Human-in-the-loop only

The local mirror and the response cache never store the secret fields those credential reads return, and allocation rows reach the local mirror with every software license key in serialOrLicense replaced by a one-way fingerprint (license: plus 16 hex characters); Firebox and access point serial numbers are kept, the row's own productName decides which; see governance.md. Through the MCP server these reads return their secrets to the agent like any other result.

The strongest control is which Access ID you give it: with the read-only Access ID, WatchGuard refuses every write. Set WATCHGUARD_MCP_READ_ONLY=1 to run the MCP server read-only. Full details, including how to lock it down, are in governance.md.

Status

Beta. Built from WatchGuard's published API specifications and tested against a local mock; no MSP has run it against a live WatchGuard Cloud account yet. Bring it to a weekly Build Session to be the first: compoundingteams.com/build-sessions.

---

Standards. Conforms to the open Agent Skills spec (Anthropic, Dec 2025; 40+ agents). MCP-compatible - works with any MCP-capable agent including Hermes. OpenClaw-ready (frontmatter pre-wired, awaiting OpenClaw launch).

Maintained by Servosity. Apache-2.0 licensed. Built with CLI Printing Press. Last updated: 2026-10-15.