excalidraw-diagram-agent
Generates editable Excalidraw diagram files from natural language descriptions of systems and architectures.
README
Built as a portfolio project proving MCP breadth beyond data-query tools: this server takes a described system and produces a real, editable diagram file — not a data lookup, a generative/creative tool exposed the same way as the data-query MCP server in this portfolio.
excalidraw-diagram-agent
An MCP server that turns a described system into a real, editable
.excalidraw diagram file. Describe an architecture in Claude Desktop
("draw me a diagram of a client talking to an API gateway, which routes to
an auth service and an order service, both writing to Postgres") and get
back a file you can open and keep editing in Excalidraw.
What's here
layout.py— a simple layered graph layout (no external graph library): positions nodes left-to-right by distance from a root, stacked vertically within each layer.excalidraw_gen.py— builds valid.excalidrawJSON: rectangles with properly bound text labels (they move together in the editor) and arrows with realstartBinding/endBinding(they stay attached when you drag a box) — not just static lines dropped at fixed coordinates.server.py— the MCP server, one tool:generate_diagram(title, nodes, edges).requirements.txt— one dependency, pinned.
1. Setup
cd excalidraw-diagram-agent
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
2. Test it standalone (no Claude needed yet)
python3 -c "
from server import generate_diagram
print(generate_diagram(
title='Order Processing System',
nodes=[
{'id': 'client', 'label': 'Web Client'},
{'id': 'gateway', 'label': 'API Gateway'},
{'id': 'auth', 'label': 'Auth Service'},
{'id': 'orders', 'label': 'Order Service'},
{'id': 'db', 'label': 'Postgres DB'},
],
edges=[
{'from': 'client', 'to': 'gateway'},
{'from': 'gateway', 'to': 'auth'},
{'from': 'gateway', 'to': 'orders'},
{'from': 'auth', 'to': 'db'},
{'from': 'orders', 'to': 'db'},
],
))
"
Should print a confirmation with the file path. Open the generated file at
excalidraw.com (File > Open, or just drag the .excalidraw file onto
the page) to see the actual diagram.
3. Connect it to Claude Desktop
Same pattern as the rental-data MCP server (project 1) — add a second
entry to the same config file, ~/Library/Application Support/Claude/claude_desktop_config.json,
alongside your existing mcpServers block:
"diagram-agent": {
"command": "/absolute/path/to/excalidraw-diagram-agent/venv/bin/python3",
"args": ["/absolute/path/to/excalidraw-diagram-agent/server.py"]
}
(Both servers can live in the same mcpServers object — don't replace
rental-data, add diagram-agent next to it.) Fully quit and reopen
Claude Desktop, then check Settings > Developer to confirm both show as
running.
4. Demo prompts
- "Draw me a diagram of a typical 3-tier web app: client, API server, database"
- "Diagram a CI/CD pipeline: developer pushes to Git, triggers a build, runs tests, then deploys to staging and production"
- "Show me an event-driven architecture with a producer, a message queue, and two consumers"
Claude will figure out the nodes and edges from your description and call the tool — you're not writing the JSON yourself, that's the point. Open the resulting file at excalidraw.com to show it's a real, editable diagram, not a static image.
Architecture notes
- Why a real Excalidraw file, not just a picture: an LLM could
describe a diagram in words, or even generate an SVG. Producing an
actual
.excalidrawfile with correct element bindings means the output is a genuine editable artifact — drag a box, the arrows and label follow. That's a meaningfully different (and harder) target than a static image. - Layout: intentionally simple (layered left-to-right) rather than a general-purpose graph layout library — sufficient for the kind of system/architecture diagrams this tool is aimed at, and keeps the dependency list at just the MCP SDK.
- Validation:
server.pychecks that every edge references a real node id before generating anything, returning a clear error instead of producing a broken file.
Repo structure
excalidraw-diagram-agent/
├── README.md
├── requirements.txt
├── layout.py
├── excalidraw_gen.py
├── server.py
└── outputs/ (generated .excalidraw files, gitignored)
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.