mcp_excalidraw and mcp-excalidraw-local
These are a fork relationship where B extends A with enterprise features (multi-tenant isolation, SQLite persistence, versioning, full-text search) while maintaining the same underlying Excalidraw MCP server architecture, making them alternatives rather than complements.
About mcp_excalidraw
yctimlin/mcp_excalidraw
MCP server and Claude Code skill for Excalidraw — programmatic canvas toolkit to create, edit, and export diagrams via AI agents with real-time canvas sync.
Implements a persistent, programmatic canvas with 26 MCP tools for element-level CRUD operations, enabling AI agents to iteratively inspect and refine diagrams via `describe_scene` and `get_canvas_screenshot`. Syncs in real-time over WebSocket to a standalone Excalidraw instance, with fallback to REST API for environments without MCP support. Supports multi-agent concurrent editing, layout automation (`align_elements`, `distribute_elements`), file I/O, snapshots, and Mermaid conversion — differentiating it from stateless, one-shot diagram generation approaches.
About mcp-excalidraw-local
sanjibdevnathlabs/mcp-excalidraw-local
Excalidraw MCP server with multi-tenant workspace isolation, SQLite persistence, auto-save, 32 tools, element versioning, full-text search, project management, and embedded canvas lifecycle. Fork of yctimlin/mcp_excalidraw with production-grade enhancements for multi-workspace Cursor/Claude setups.
Implements a stdio-based MCP server that embeds a WebSocket-synced Excalidraw canvas, using better-sqlite3 for versioned element storage and automatic workspace detection via environment variables. Includes 32 domain-specific tools (shape creation, Mermaid→Excalidraw conversion, image embedding, element search) alongside an optional agent skill with playbooks and helper scripts for Cursor/Claude integration.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work