Support LLM Discovery and Usage of Prompts and Resources

[kartheekp-ms]

Is your feature request related to a problem? Please describe.

Currently, Prompts are explicitly user-controlled and Resources are application-driven, which limits the flexibility of LLMs in dynamically discovering and utilizing them. This can lead to missed opportunities for automation, context-awareness, and intelligent suggestions, especially in complex workflows where the model could benefit from proactively surfacing relevant prompts or resources.

Reference https://modelcontextprotocol.io/docs/learn/server-concepts#core-server-features

Describe the solution you'd like

I’d like to propose making Prompts and Resources optionally LLM-controlled, similar to how Tools are handled. This would allow models to:

• Discover and suggest relevant prompts based on context.
• Dynamically retrieve and include resources to enrich responses.
• Use parameter completion and metadata to guide users intelligently.
• Maintain user oversight through approval mechanisms, visibility settings, and activity logs.

This hybrid control model would preserve user agency while unlocking more powerful and adaptive workflows.

Describe alternatives you've considered

• None as of now.

Additional context

By extending LLM control to prompts and resources, we can enable richer interactions like:

• Auto-suggesting prompts in context menus or command palettes.
• Pre-fetching resources based on conversation history or inferred needs.
• Creating adaptive workflows that evolve with user input and model reasoning.

Similar ask: langchain-ai/langchain-mcp-adapters#62