[Feature Request] Expose Claude Design generative capabilities via MCP methods

Bug Description
Feature Request: Expose Claude Design's generative capability via MCP (not just file sync) Summary The current Claude Design MCP (DesignSync) only supports file-level operations - read/write/CRUD of project files. It does NOT expose the generative design capability that powers the claude.ai/design web product. As a result, an agent (e.g. Claude Code) cannot programmatically initiate or iterate a design from a natural-language brief. Current limitation Available methods are limited to list_projects / get_project / list_files / get_file / create_project / write_files / delete_files. There is no generate / iterate method. To generate a design, a human must manually drive the web UI; the agent can only pull the result afterward. Requested capability An MCP method such as generate_design(projectId, prompt, targetFiles?, referenceFiles?) and iterate_design(projectId, path, instruction) that triggers the same generative agent used by claude.ai/design, so a coding agent can: 1. Send a design brief and receive generated screens/components into a project. 2. Iterate on specific files with follow-up instructions. 3. Close the loop end-to-end: generate design -> pull -> implement in the real repo, all from one agent session. Use case I use Claude Code to maintain a production app and claude.ai/design to explore visual direction. Today I must context-switch to the web UI to generate, then return to Claude Code to implement. A generative MCP method would let the agent own the full "design -> code" loop, which is exactly the promised value of bridging design and code.

Environment Info

• Platform: darwin
• Terminal: iTerm.app
• Version: 2.1.198
• Feedback ID: c9fa2a6a-84ba-482d-bfe7-a4bad7244c4a

Errors

[]