For use with agentic harnesses that don't support MCP apps, there's no good cycle to create an image, view it, and edit it. The only thing the AI can do is export it to excalidraw, but then if further modifications are needed, it can't do anything about it. Instead, can you do something like live collaboration mode on the website, where the MCP server starts a collaboration, invites the user (so they can open it in their browser), and then edits can continue from there?
For use with agentic harnesses that don't support MCP apps, there's no good cycle to create an image, view it, and edit it. The only thing the AI can do is export it to excalidraw, but then if further modifications are needed, it can't do anything about it. Instead, can you do something like live collaboration mode on the website, where the MCP server starts a collaboration, invites the user (so they can open it in their browser), and then edits can continue from there?