Existing XMind MCP tools make LLMs output the entire mind map as one giant JSON blob. That's wasteful and error-prone. I built an MCP server with 19 atomic tool

Existing XMind MCP tools make LLMs output the entire mind map as one giant JSON blob. That's wasteful and error-prone. I built an MCP server with 19 atomic tools instead -- add_topic, move_topic, delete_topic, etc. -- so LLMs can make surgical edits using stable node IDs. Python/FastMCP, 70+ tests, reverse-engineered from XMind's official generator since there's no public spec.