initial mcp prompt support #11
Description
this feature would map mcp PromptMessage results to either HumanMessage or AIMessage, similar to what is already done for tools. this allows users to pass prompts retrieved from an mcp server directly to langchain llm invocations
initial support would handle fetching prompts containing text/image content
however, i'm not sure how we should handle embedded resources. i'm thinking resources can mapped to HumanMessage but there's some work to do to support the different mime types we might encounter
perhaps for now we raise a "not implemented" error when embedded resources are encountered, and add support when this package has general resource support
i'm happy to work on this pending feedback from the community