Add Fleet MCP server to community servers

[SimplyMinimal]

Description

Server Details

• Server: Fleet MCP (community contribution)
• Changes to: README.md - Community Servers section

Motivation and Context

This PR adds the Fleet MCP server (https://github.com/SimplyMinimal/fleet-mcp) to the community servers list. Fleet MCP is a Model Context Protocol server that enables AI assistants to interact with Fleet Device Management for device management, security monitoring, and compliance enforcement.

The server provides capabilities including:

• Host management
• Live query execution
• Policy management
• Software inventory
• Vulnerability tracking
• MDM operations
• Read-Only and Read-Write modes

By default, adding this MCP server will allow read-only access to interact with a Fleet instance once an API key is provided.

How Has This Been Tested?

This was tested using LM Studio as the client. Tested scenarios include the following (not an exhaustive list):

• Create a policy that will do XYZ.
• Run the <script name> against host-xyz-123
• Run a one-time script against host-xyz-123 (useful if you do not want to upload a script and just want to execute a script ad-hoc)
• Add a printer to host-xyz-123
• List the teams and their members in Fleet
• Which servers are affected by CVE-XXXX
• Which machines fail the disk encryption policy?

To experiment with this against a dummy instance of Fleet, you can use fleetctl preview but note that this will default to a non-Premium instance of Fleet which may limit the options available.

Breaking Changes

No breaking changes. This is a documentation-only update for a new MCP server addition.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update

Checklist

• I have read the MCP Protocol Documentation
• My changes follows MCP security best practices
• I have updated the server's README accordingly
• I have tested this with an LLM client
• My code follows the repository's style guidelines
• New and existing tests pass locally
• I have added appropriate error handling
• I have documented all environment variables and configuration options

Additional context

Recommendations:

• This MCP server is dependent on a good system prompt to make the most out of the server. This is being worked on to automatically suggest the best system prompt context but for now this must be manually loaded for better performance.
• This is an experimental approach to interacting with Fleet so although the interactions should be safe, the accuracy of the data should always be reviewed. As in, never trust an LLM to be 100% accurate but you can use this to get you started in the right direction. For example, this is super helpful when trying to come up with a new query or policy without a deep understanding of the various tables (100+ tables) that are available in Fleet.
• Another recommendation is to set the temperature to be low to reduce hallucinations.
• If you plan to enable write mode (the server defaults to read-only), it can help with creating scripts and policies among many other changes when prompted. However, it's best to have your client prompt for approval/confirmation in those scenarios. LM Studio offers a capability where if a tool is a read operation it will be allowed automatically whereas a tool that involves writing will prompt before proceeding. Whichever client you use, it's best to adopt this behavior.
• Here's a full list of available tools: https://github.com/SimplyMinimal/fleet-mcp#available-tools

[olaservo]

Thanks for your contribution to the servers list. This has been merged in this combined PR: #2973