fix(client): Support specifying windowsHide for STDIO servers

[erran]

Motivation and Context

Previously windowsHide was only available when running inside of an Electron process by default. If this library is run through a single application executable the NodeJS runtime will cause windowsHide to be set to false thus spawning terminals for each spawned STDIO MCP server.

This is obtrusive and in the foreground of the user’s IDE in my experience with little benefit due to the STDOUT/STDERR being hidden from the terminal window as it is captured by the NodeJS process.

How Has This Been Tested?

1. New unit test coverage.
2. Tested stdio MCP servers before and after these changes were applied by a patch-package post install script in the GitLab Language Server which is used as a standalone binary outside of electron.

Breaking Changes

n/a — the stdio transport will behave as it does today without explicitly setting the windowHide option.

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 Documentation
• My code follows the repository's style guidelines
• New and existing tests pass locally
• I have added appropriate error handling
• I have added or updated documentation as needed

Additional context

Usage of these changes at GitLab to avoid opening terminal windows when running under Windows environments:

1. https://gitlab.com/gitlab-org/editor-extensions/gitlab-lsp/-/merge_requests/2533

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/modelcontextprotocol/typescript-sdk/@modelcontextprotocol/sdk@1104

commit: 98f7842

[erran]

Honestly I was debating whether hard coding true here or removing the electron check was sufficient due to how it’ll behave on non-Windows systems but I didn’t get around to testing different platforms since I’m usually developing on Mac or Linux so validating the changes in windows took most of my day.

I decided against true or win32 platform check as the following code could break someone else’s assumptions in a way I’m not thinking of right now (although in my experience the terminal window for each stdio MCP server is always blank).

Suggested change

	windowsHide: this._serverParams.windowsHide ?? isElectron(),
	windowsHide: process.platform === 'win32',

[erran]

I have a single test failing on main locally but it is in a different area of the codebase. I can paste details here if there aren’t any known intermittent failures relating to sse transport on an ARM Mac and receiving too many events.