Draw.io Scientific Illustrator provides live, visible control of the draw.io canvas via MCP servers and a Codex plugin to draw figures step-by-step. It includes installation scripts, environment configuration, and live workflow commands.
What it is
Draw.io Scientific Illustrator is a Codex plugin that lets an AI agent draw scientific figures live inside the visible draw.io desktop canvas. It uses a localhost-only MCP server to call draw.io's graph API and does not automate OS-level mouse/keyboard or pre-create XML. The project includes MCP servers for live drawing and file utilities, plus a Codex skill and a plugin marketplace configuration for installation.
How it works
The repository contains:
- drawio-live MCP server: launches/connects to the visible draw.io desktop editor and edits its active graph model in real time.
- drawio-file-utils MCP server: validates saved .drawio documents and exports PNG, SVG, PDF, or JPG deliverables.
- A Codex skill to guide the agent in inspecting references, decompose into editable primitives, pace drawing, review, refine, and save post-visible drawing.
- A repository-local Codex marketplace for installing the complete plugin as a unit.
The workflow for live operation follows a sequence: launch or connect to the visible editor, confirm graph readiness, add shapes/edges (paced), screenshot and inspect, update cells, fit view, save a snapshot, then validate and export.
Getting started
Installation options include:
- Install — easiest methods: paste the provided Codex task text to install from GitHub, clone, register marketplace, and install drawio-scientific-illustrator@drawio-scientific-tools.
- Windows one-command installer: a PowerShell command generating and running a script from install.ps1.
- macOS/Linux one-command installer: a curl | bash command running install.sh.
- Manual install: clone the repo, add to Codex marketplace, and install the plugin using codex commands.
The repository provides exact command blocks for installation in both English and Chinese.
Getting started (usage and configuration)
Usage flow includes restarting Codex, then attaching a reference page (PNG/JPEG/SVG or PDF page), selecting the Draw.io Scientific Illustrator in Codex, and specifying pacing and output formats. The recommended Codex configuration for complex redraws mentions selecting GPT-5.6 Sol and enabling a six-level reasoning selector to access the Max option.
Configuration options include environment variables:
- DRAWIO_PATH: Explicit draw.io executable path (default auto-detected)
- DRAWIO_LIVE_PORT: Preferred localhost debugging port (default 9333; nearby free port allocated if needed)
- DRAWIO_LIVE_PROFILE: draw.io/Electron profile directory (default ~/.drawio-live-mcp/<port>)
Example usage shows how to set DRAWIO_PATH on Windows.
Recent releases
Latest release is v1.0.0 Draw.io Scientific Illustrator v1.0.0 (2026-07-12): First public release. Live, visible draw.io canvas control through MCP; bilingual English/Chinese installation and usage guides; Windows-tested automatic draw.io discovery; validation, export, install.
Traction
As of the repository facts, stars: 1086 and stars_1d: 254. Forks: 74. Open issues: 3. The README confirms Windows as tested for v1.0.0 and cross-platform considerations, but no further trend data beyond the raw numbers is provided here.
Behind the repo
Not applicable: no linked startup or company page provided beyond the GitHub repository.
Caveats
- License: MIT
- Created: 2026-07-12
- Last push: 2026-07-21
- Known limitations include: Windows is the tested platform for v1.0.0; macOS/Linux support is best effort; live behavior may vary with Electron packaging. The security notes mention the endpoint binds to 127.0.0.1 and that no telemetry or hosted backend is included.
- Requires Node.js 22+ for MCP servers running outside the Codex runtime.
License
MIT © 2026 icebird1998