Codify MCP tool name requirements into the spec

[domfarolino]

MCP has official constraints on what a tool name can and cannot contain: https://modelcontextprotocol.io/specification/2025-11-25/server/tools#tool-names. These are SHOULDs, not MUSTs, but I think we should carry over these the best practices into the platform, since it increases developer feedback and predictability, and decreases subtle errors.

One of the motivations for this was trying to write a WebMCP tool in Chrome whose name contained a space, and scratching my head when built-in Gemini failed to call it, with no good feedback. As a developer, I would've appreciated the platform catching this and giving some feedback on my tool definition, instead of implicitly loading it into the model when it violates something so simple.

[beaufortfrancois]

Do you expect registerTool({name: "name with space", ...}) calls and toolname="name with space" declarations to raise a warning in browser DevTools?

[DavidMulder0]

This makes sense if an alternative display name property is added to tools (see #133 ). (At the moment WebMCP more closely mirrors an outdated MCP spec without title that contained no such restrictions on name: https://modelcontextprotocol.io/specification/2025-03-26/server/tools#tool)