lgazo/drawio-mcp-server

What it solves

This project provides a way for AI agents to programmatically create, edit, and manage diagrams in Draw.io (Diagrams.net). It eliminates the need for users to manually draw architectural diagrams, flowcharts, and visual documentation by allowing an AI assistant to handle the layout and element placement.

How it works

It operates as a Model Context Protocol (MCP) server. It can either host a built-in Draw.io editor in the browser or connect to an existing Draw.io session via a browser extension. The server exposes a set of tools that AI agents can use to add shapes, manage layers, create pages, and handle document persistence (saving and titling).

Who it’s for

Developers, architects, and technical writers who use MCP-compatible AI clients (like Claude Desktop, Claude Code, or Zed) and want to automate the visual documentation of their systems.

Highlights

• Multi-modal targeting: Supports multi-document and multi-page workflows, allowing agents to work across different tabs or pages.
• Vendor shape support: Automatically discovers stencils for AWS, GCP, Azure, and Cisco, enabling agents to use industry-standard icons.
• Flexible deployment: Offers both a built-in editor and a browser extension for integration.
• Diagramming tools: Includes capabilities for importing/exporting XML, SVG, and PNG, as well as Mermaid diagram integration.
• Concurrency safety: Uses per-document FIFO serialization to ensure multiple agents can modify different files safely.