About staruml-mcp
staruml-mcp lets AI agents control StarUML programmatically over MCP. It is for developers generating or managing UML diagrams from agents; Mermaid-based diagram tools are built in, with project and model operations available through an extension.
Key features
- 01Generate UML diagrams from Mermaid
- 02Inspect diagrams and export PNG images
- 03Manage project files through the extension
- 04Find and execute StarUML commands
- 05Create, query, update, and delete model elements
Use cases
- 01Generate UML diagrams from descriptions
- 02Inspect diagrams in the current project
- 03Update model elements through an agent
MCP support
As an MCP server, it exposes StarUML tools to agents over stdio or HTTP and communicates with StarUML through its HTTP API. Some project, command, and element-management tools require the separate staruml-mcp-extension.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
19Saved tool list · updated 3w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| code | string | Yes | Mermaid diagram source code. Must start with one of: classDiagram, sequenceDiagram, flowchart, erDiagram, mindmap, requirementDiagram, stateDiagram. Example: "flowchart LR\n A[Start] --> B[End]" |
Details
From the project READMEstaruml-mcp
Model Context Protocol (MCP) server for StarUML. Lets AI agents (Claude Code, Cursor, VS Code Copilot, Codex) drive StarUML programmatically — generate UML diagrams from Mermaid, execute any built-in command, CRUD elements, save projects, and more.
How it fits together
AI Agent ──MCP──► staruml-mcp (this package) ──HTTP──► StarUML
:58321 (built-in, 4 tools)
:58322 (extension, 15 tools)
| Package | What it is | Where it runs |
|---|---|---|
staruml-mcp (this repo) | MCP server for AI agents | your machine via npx -y staruml-mcp |
staruml-mcp-extension | StarUML plugin adding 15 HTTP endpoints | inside StarUML (install once via Extension Manager) |
- Using only Mermaid-based diagram tools? Install
staruml-mcponly. The 4 built-in tools work. - Want the full 19 tools (project save/open, element CRUD, execute any StarUML command)? Install both.
Prerequisites
- StarUML v7.0.0+ with API Server enabled (see below)
- Node.js 20+ on the machine running the AI agent
- (Optional)
staruml-mcp-extensioninstalled in StarUML — required for 15 of the 19 tools
Enable StarUML API Server
Edit settings.json at:
- Windows:
%APPDATA%\StarUML\settings.json - macOS:
~/Library/Application Support/StarUML/settings.json - Linux:
~/.config/StarUML/settings.json
Add or update:
{
"apiServer": true,
"apiServerPort": 58321
}
Restart StarUML.
Verify:
curl http://localhost:58321/
# → "Hello from StarUML API Server!"
Install & Use
Claude Code (recommended — HTTP transport)
Start the server in a terminal:
npx -y staruml-mcp --transport http
# listens on http://localhost:58323/mcp by default
Register with Claude Code:
claude mcp add --transport http staruml http://localhost:58323/mcp
Port
58323is the canonical HTTP port, chosen to sit alongside StarUML's built-in API (58321) andstaruml-mcp-extension(58322). Override with--port <n>if needed.
Restart Claude Code. Ask:
"What StarUML tools do you have?"
Claude Desktop (stdio transport)
Edit %APPDATA%\Claude\claude_desktop_config.json (Windows) or ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):
{
"mcpServers": {
"staruml": {
"command": "npx",
"args": ["-y", "staruml-mcp"]
}
}
}
Restart Claude Desktop.
Cursor / VS Code Copilot / Codex CLI
Point your MCP client at npx -y staruml-mcp (stdio) or http://localhost:58323/mcp (HTTP).
CLI
staruml-mcp [options]
-t, --transport <type> stdio | http (default: stdio)
-p, --port <number> HTTP listen port (default: 58323)
--api-port <number> StarUML built-in API port (default: 58321)
--ext-port <number> staruml-mcp-extension port(default: 58322)
--api-host <url> StarUML API host prefix (default: http://localhost)
-V, --version Print version
-h, --help Show help
Tools Exposed
Built-in (always available, port 58321)
| Tool | Description |
|---|---|
generate_diagram | Generate a UML diagram from Mermaid code. |
get_all_diagrams_info | List all diagrams in the current project (id, name, type). |
get_current_diagram_info | Get metadata of the currently focused diagram. |
get_diagram_image_by_id | Export a diagram as PNG by its ID. |
Extension tools (require staruml-mcp-extension, port 58322)
| Tool | Description |
|---|---|
get_all_commands | List all 138+ built-in StarUML command IDs. |
execute_command | Execute any built-in command (e.g. project:save, view:fit-to-window). |
get_project_info | Current project's filename + top-level elements. |
save_project / save_project_as / new_project / open_project | Project file lifecycle. |
get_element_by_id / find_elements | Query model elements. |
create_element / update_element / delete_element | Element CRUD. |
create_diagram (typed native) / switch_diagram / close_diagram | Diagram management. |
To enable extension tools: install staruml-mcp-extension in StarUML (Tools → Extension Manager → Install From URL → https://github.com/ezrabrilliant/staruml-mcp-extension).
Example Prompts
- "Create an ER diagram in StarUML for a POS database: users, menus, transactions with relationships."
- "Generate a sequence diagram for JWT login: frontend → /api/auth/login → AuthService → DB → JWT response."
- "Show me the current diagram in StarUML."
- "Export diagram with ID
xyz123as an image."
Development
git clone https://github.com/ezrabrilliant/staruml-mcp.git
cd staruml-mcp
npm install
npm run dev # tsx watch on src/
npm run build # bundle to dist/
npm test # vitest
npm run typecheck # tsc --noEmit
Architecture
AI Agent (Claude Code / Cursor / VS Code / …)
│
│ MCP (stdio or Streamable HTTP)
▼
staruml-mcp (this package)
│
│ HTTP JSON-RPC
▼
StarUML API Server (port 58321)
│
▼
StarUML Application (v7+)
Acknowledgments
Inspired by staruml/staruml-mcp-server (official stdio-only server by Minkyu Lee, StarUML creator). This project reimplements it with multi-transport support and strict TypeScript.
License
MIT © Ezra Brilliant Konterliem
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.