MCP UI: Show rendered UI components without raw tool input/output expanded?

[notfoundry]

Every tool call in LibreChat has a chevron that, when expanded, shows the raw JSON input/output for the tool. This makes sense for standard tool calls, since the tool output is meant for machine consumption as a message during AI inference, not for direct human use. But when a MCP tool call returns a renderable MCP UI component, that output is explicitly designed for the user to see / interact with / act on. Hiding this UI behind the same chevron used for raw JSON creates a cognitive mismatch IMO.

• Users associate that chevron with developer-facing debug info in 99% of MCP tool calls, not primary interface content
• If the MCP UI is stuck in there and collapsed by default, many users will never find it, or will assume it’s just another technical detail they can ignore

I think it might be a positive change if the rendered MCP UI from a tool call was visible by default, regardless of whether the raw tool input/output panel is collapsed. Maybe the JSON should stay hidden by default, and have the chevron show/hide the same content as for every other tool call, but the rendered UI component panel after the raw tool output is the actual product of the tool call in this context. It's especially important when the UI includes interactive elements, forms, tables, or streaming updates that users need to see immediately to continue the conversation meaningfully.

Happy to submit a PR, but I want to bring it up first to understand whether this is going in the wrong direction for MCP UI support