Markasso
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
User Manual · Guide · MCP for AI agents · Architecture · Contributing · Report an Issue
Featured in
TypeScript.fm — Episode 87 · October 29, 2026
Markasso was featured in Library Watch. Listen at 25:49.
Thanks to Erik Onarheim and Kamran Ayub for the shout-out!
Related MCP server: antigravity-drawio-mcp
Draw with AI agents (MCP)
Markasso ships an MCP server: Claude, Codex, Gemini and local models (e.g. Gemma in LM Studio) can create, read and edit diagrams, and even draw live on a shared board.
1. Connect your agent (Node 22+):
# Claude Code: the plugin adds the server and a skill that knows when to draw
claude plugin marketplace add AlbertoBarrago/markasso-plugin
claude plugin install markasso@markasso
# Claude Code, server only
claude mcp add markasso -- npx -y @markasso/mcp@latest
# Codex CLI
codex mcp add markasso -- npx -y @markasso/mcp@latestClaude Desktop, Gemini CLI and LM Studio use the same npx -y @markasso/mcp@latest command in their mcpServers config: see mcp/README.md for every client, the tool list, building from source and the security notes.
2. Ask for a diagram, for example: "Draw the architecture of a checkout service with icons, lay it out left to right and give me a share link". For live mode, start a session with ⋯ More tools → Start live session in Markasso, then ⋯ More tools → Invite AI agent copies a prompt that brings the agent onto the board. Menu → Connect AI agent shows the setup for every client.
Project Status
Markasso is feature-complete. Active development of new features is paused. The tool does what it was designed to do. Bug fixes, performance improvements, and small refinements are welcome. Large feature PRs are unlikely to be merged; open an issue first to discuss.
Why Markasso
Excalidraw excels at freehand sketching and draw.io at comprehensive diagramming. Markasso occupies a different niche: smaller, faster, keyboard-first, with no sign-in, cookies or workspace to create. This repository is the source of truth for the production deployment.
It started as a game: a project small enough to finish and complex enough to be honest, to stay sharp with LLMs by actually using them. A whiteboard from scratch, vanilla TypeScript, zero dependencies. It didn't stay small. Colleagues started using it, then remote teammates, then people I've never met, to sketch ideas, run retrospectives and think out loud together.
The core (scene, reducer, renderer, tools) stays dependency-free. A dependency is fine only for an isolated, lazy-loaded feature where building it ourselves would mean re-solving someone else's solved problem, worse: Mermaid import runs on @excalidraw/mermaid-to-excalidraw and costs nothing until first use.
Features
Drawing: rectangle, rhombus, ellipse, lines with UML-ready arrowheads, curves, freehand pen with stylus pressure, text, sticky notes, shape and connector labels
Smart connections: arrows attach to shape borders and ports, follow moves, resizes and rotations, with elbow routing
Diagram libraries: flowchart shapes, generic architecture icons, the Kubernetes icon set, and official AWS, Azure and Google Cloud packs imported from their ZIP (kept local, never redistributed)
Realtime collaboration: live sessions over a Cloudflare Durable Object, with remote cursors and shared undo
AI agents: an MCP server lets agents build diagrams or draw live on a shared board (see above)
Import & export:
.markassofiles, Mermaid (flowchart, sequence, class, state, ER as native elements), images, PNG, SVG, standalone HTML, share links encoded in the URLKeyboard-first: command palette (
Ctrl+K), element search (Ctrl+F), one-key tools, hide all panels with\Canvas: infinite pan and zoom, dot, line and millimeter grids, minimap, dark and light themes, auto-save to
localStorage
Every tool, option and shortcut is documented in the User Manual (see the keyboard shortcut reference).
Getting Started
Install
pnpm installDevelop
pnpm dev # Start dev server at http://localhost:5173Build & Test
pnpm build # Type-check and bundle into dist/
pnpm typecheck # TypeScript validation
pnpm test # Run Vitest unit tests
pnpm mcp:build # Build the MCP server → mcp/dist/markasso-mcp.jsNo .env files. No API keys. No containerization required. Cloud boards and licensing are off in local builds (VITE_CLOUD_ENABLED).
pnpm coverage runs the suite and updates the coverage badge. UI and rendering modules need a browser and are excluded from Node tests by design. How the code is organized is described in ARCHITECTURE.md.
Contributing
See CONTRIBUTORS.md for guidelines.
License
MIT · © 2026 Alberto Barrago. Bundled third-party material (Kubernetes icons, Apache-2.0) is listed in THIRD_PARTY_NOTICES.md.
Draw things. Ship things. Touch grass.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Real-time collaborative whiteboard — AI agents and humans edit the same board live over MCP.
Collaborative whiteboard MCP server — create objects, connectors, C4 diagrams, and manage boards
MCP server that lets your AI assistant build and deploy apps, hosting each one at a live URL.
One memory, every AI. A shared, user-owned markdown memory your AI clients read and write over MCP.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceAn MCP server that gives AI agents programmatic control over a live tldraw canvas, enabling diagram creation through natural language.1MIT
- AlicenseBqualityCmaintenanceMCP server that enables AI assistants to create, parse, render, and validate Draw.io diagrams programmatically.1473 PyPI2MIT
- FlicenseNot gradedqualityCmaintenanceMCP server for a persistent local whiteboard that lets AI assistants read, add, update, and delete canvas elements, with undo/redo and provenance tracking.-
- FlicenseNot gradedqualityBmaintenanceMCP server for Escalidrau, a collaborative macOS whiteboard. Lets AI agents read, draw, edit, rearrange, and export diagrams on a shared canvas in real time.1-