
Windows Server MCP
io.github.ServosityPublished by GitHub account Servosityv0.1.0Updated Nov 7, 2026
Agentless Windows Server health checks over WinRM: one verdict per server, a diffable inventory
Overview
Agentless Windows Server health checks over WinRM: one verdict per server, a diffable inventory, and no arbitrary PowerShell.
- What you can do
- Check health and inventory across all configured servers, with history and diffs between snapshots
- Inspect services, event logs with baselines, certificates, scheduled tasks and local admin changes
- Read system and volume details from a single named server
- Restart or start a service after a preview, only with explicit confirmation
- When to use it
- For MSP or IT staff who want an assistant to answer which Windows servers need attention today, what changed since the last snapshot, and why a service stopped, without opening an RDP session per server. It complements, rather than replaces, an RMM.
- Requirements
- Runs as a local binary (stdio) on the user's machine, or over HTTP with a bearer token for remote agents. Needs WinRM over HTTPS (port 5986) enabled on each target server, an account in that server's Administrators group, and a config file listing servers with certificate pinning or a CA file. Credentials come from WINDOWS_SERVER_USERNAME and WINDOWS_SERVER_PASSWORD or a per-server login.
Installation
In SourceWeft
- Open Windows Server MCP in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
Windows Server + 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 Windows Server. Not affiliated with, endorsed by, or sponsored by Microsoft Corporation.
[Windows Server demo - animated preview]
▶ Watch the 30-second demo - demo data is simulated; every command shown exists in the real CLI.
Agentless Windows Server health checks over WinRM: one verdict per server, a diffable inventory, and no arbitrary PowerShell. Works with the AI you already use - ChatGPT (Plus/Pro+), 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
The six agents MSP owners actually use (self-serve, works today):
For ChatGPT, run windows-server-mcp --transport http with WINDOWS_SERVER_MCP_HTTP_TOKEN set, on an HTTPS host; every request sends that token. See 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.
Skill-native agents (also covered): Hermes and OpenClaw read this skill's
SKILL.mddirectly 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 Windows Server MCP (.mcpb) - then open Claude Desktop > Settings > Extensions and select the file. One click, no JSON, no shell. (Browse every Windows Server release on the releases page.)
Prefer the Claude Code plugin? Add the marketplace once, then install - works immediately, no directory listing required:
Path A - paste one prompt into your AI agent (recommended)
Copy this into Claude Code, Codex CLI, or Claude Cowork:
Install the Windows Server 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/windows-server/install.sh). If it uses Windows PowerShell, runiwr -useb https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/windows-server/install.ps1 | iex. Then authenticate per the README and runwindows-server-cli --helpto explore.
The same prompt works in any agent that can run shell.
Path B - run the installer yourself
Windows (PowerShell):
macOS / Linux:
The installer drops both windows-server-cli (the CLI) and windows-server-mcp (the MCP server) into your user bin path. Claude Code, Codex, and Cowork discover the Skill via SKILL.md in this directory.
Verify:
Upgrade to the latest version
The installer always fetches the current release - re-run it to upgrade:
macOS / Linux:
Windows (PowerShell):
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 windows-server@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:
Inside a Hermes chat session:
Hermes speaks MCP natively, so it can also use the windows-server-mcp server directly - same install path, same env vars.
Install for OpenClaw
Tell your OpenClaw agent (copy this):
Install the windows-server skill from https://github.com/servosity/msp-skills/tree/main/skills/windows-server. The skill defines how its required CLI (
windows-server-cli) can be installed via theopenclaw:frontmatter block.
OpenClaw isn't generally available yet; the frontmatter wiring is pre-shipped and will activate the moment OpenClaw launches.
Authenticate
Windows Server has no API key: the CLI signs in to each server's WinRM HTTPS listener
(port 5986) with a local or domain account in that server's Administrators group. Add
each server to the config file (default ~/.config/windows-server-cli/config.toml on
macOS and Linux, %USERPROFILE%\.config\windows-server-cli\config.toml on Windows;
override with --config or WINDOWS_SERVER_CONFIG):
Then store its password (read from stdin, kept only in the CLI's 0600 credential store) and run the real connectivity test:
macOS / Linux:
Windows (PowerShell):
Or share one account across servers with WINDOWS_SERVER_USERNAME and
WINDOWS_SERVER_PASSWORD. Passwords shorter than 8 characters are refused. With NTLM, a
local account other than the built-in Administrator gets HTTP 401 unless
LocalAccountTokenFilterPolicy is 1 on the server: use a domain account, or
auth = "basic" over HTTPS. doctor checks config and paths and reports whether
credentials are present; it does not contact a server. health is the connectivity
test.
What this skill does
health, services down, eventlog new, local-admins changes and inventory snapshot read every configured server; the single reads and the service actions act on
one server (--server <name>, or the entry marked default = true).
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
Most ways to put an AI on a Windows server hand it a PowerShell prompt. This one cannot run arbitrary script: every command is one of a fixed set of read scripts that ship inside the binary, checked by SHA-256 on the server before they run, with arguments validated and passed as typed data. The server comes from your config by name (never from the agent), certificate verification is always on, and nothing is installed on the server.
On top of the reads, compound commands do work a single PowerShell call cannot: health
rolls every check into one verdict per server and remembers how long each has failed,
services down pairs each stop with its Service Control Manager event, eventlog new
compares against the server's own history, and inventory snapshot plus
inventory diff keep a local SQLite history so you see what changed between runs.
The pain this closes
Daily server checks are still a manual round on many MSP desks. The Active Directory Pro Windows Server maintenance checklist has you "monitor critical server services" every day, review disk space weekly, and check Event Viewer's application, security and system logs "for any critical errors": an RDP session per server, per client. And when a server breaks, the first hour goes to working out what changed.
healthanswers "which servers need attention" in one call, with reasons.services downstarts a down ticket with the cause, not a restart.eventlog newseparates a new failure from the error that fires every night.local-admins changesandinventory diffname what changed, who did it, and when.
See pain-point.md for the longer narrative.
Frequently asked questions
Does this work with ChatGPT?
Yes, on Plus, Pro, Team, Business, Enterprise, and Education plans (Free tier does not yet expose Developer Mode). ChatGPT connects to remote MCP servers over HTTPS, not local stdio binaries. The Windows Server MCP server runs locally over stdio by default. Its HTTP mode (windows-server-mcp --transport http) serves http://127.0.0.1:7777/mcp, refuses to start until WINDOWS_SERVER_MCP_HTTP_TOKEN is set, requires Authorization: Bearer <token> on every request, and needs --tls-cert/--tls-key for any other address. See mcp-install.md. Step-by-step in mcp-install.md.
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 Windows Server credentials once.
Is my Windows Server data safe?
Your data stays on your machine. The CLI and MCP server are local binaries. The SQLite mirror sits in a directory under your user account. The AI agent only sees what the CLI returns - typically a query result, not raw bulk data. Credentials are read from your environment or your agent's config; never bundled into this repo or transmitted anywhere by MSP Skills.
Do I need to install anything on the server?
No. It uses WinRM over HTTPS (port 5986), which is built into Windows Server. Enable an HTTPS listener (Microsoft's guide: Installation and configuration for Windows Remote Management) and give the CLI an account in the server's Administrators group.
Can the AI run its own PowerShell on my client's server?
No. Every command runs one fixed script that ships inside the binary. A service name or log name is checked against a strict pattern and passed as typed data; input that would change the script is refused.
What can it change on a server?
Two things, each after a preview: restart a service (services restart) or start a
stopped automatic service (services start). The CLI needs --yes; the MCP tools need
confirm: true. WINDOWS_SERVER_MCP_READ_ONLY=1 removes both from the MCP server.
Does it replace my RMM?
No. It complements it: an RMM agent alerts on thresholds; this answers "why" and "what changed" questions on demand, across every server in your config.
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
The strongest control is the account you give it: the CLI can only do what that account may do on the server, and only through its fixed scripts. Full details, including how to lock it down, are in governance.md.
Status
Beta. Verified live against a Windows Server 2022 test server we run, and being validated with MSPs running it live against their own client servers in our weekly Build Sessions. RSVP at 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: 2026-10-29.
Source: skills/windows-server/README.md at commit 62ab9f0
Tools
0Version history
1- v0.1.0LatestNov 7, 2026
Score
40% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts
Declare a license in the repository or the registry entry.

