ParentFull thread_pdp_·Very cool. Why not start with an MCP instead?An existing LLM could drive the generation while the MCP can render the final result?View on HN