# Diagrams MCP [Health: Active]

**Category:** 📊 Data Visualization  
**Repository:** https://github.com/ByteOverDev/diagrams-mcp  
**GitHub Stars:** 3  
**Views:** 2  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/diagrams-mcp

## Description
Generate cloud architecture diagrams, flowcharts, and sequence diagrams.

## Tools
Capabilities this server exposes over MCP:

- **aws**
- **gcp**
- **k8s**
- **azure**
- **onprem**
- **compute**
- **database**
- **network**
- **Image**
- **dict**
- **EC2**
- **ComputeEngine**

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "diagrams-mcp": {
    "command": "uvx",
    "args": ["diagrams-mcp-server"]
  }
}
```

## Documentation

## What the Diagrams MCP MCP server does

Diagrams MCP generates diagram images from an MCP client. It supports cloud architecture diagrams through the Python-based mingrammer/diagrams library, as well as general diagram definitions written for Mermaid and PlantUML. The available output includes infrastructure layouts, flowcharts, sequence diagrams, class diagrams, entity-relationship diagrams, state diagrams, Gantt charts, deployment diagrams, and other formats supported by the selected engine.

The Diagrams MCP MCP server can work with providers and service categories such as AWS, GCP, Azure, Kubernetes, and on-premises infrastructure. Its discovery operations help an agent identify provider names, service groups, node classes, and import paths before it creates a cloud diagram. Cross-provider lookups can also map an infrastructure node to an equivalent service in another provider, such as AWS EC2 and GCP ComputeEngine.

## How it works

Cloud diagrams are generated by executing diagrams-library Python code in a subprocess and returning a rendered image. Mermaid definitions are processed separately and can produce PNG or SVG output. PlantUML definitions produce PNG output and cover UML-oriented diagram types.

The server also exposes MCP resources containing reference material for diagram constructors, edges, clusters, Mermaid syntax, and PlantUML syntax. These resources give an agent examples and parameter guidance without requiring the information to be embedded in every prompt.

For hosted deployments, the MCP process can run as a facade and send rendering work to a separate HTTP renderer service. This arrangement keeps Graphviz, Chromium, Mermaid CLI, Java, and PlantUML in the renderer environment rather than the always-on facade process.

## Setup and configuration

The local stdio installation can use `uvx diagrams-mcp-server`, a package installed with pip, or a source checkout. Graphviz is needed for local cloud architecture rendering. Mermaid CLI is needed for local Mermaid rendering, while PlantUML requires Java and the PlantUML JAR. These dependencies are engine-specific rather than all being required for every use case.

Claude Desktop, Claude Code, Cursor, Windsurf, and VS Code can launch the server with the documented stdio configurations. Remote facade mode uses `FASTMCP_TRANSPORT`, `FASTMCP_HOST`, and `FASTMCP_PORT`, together with `DIAGRAMS_RENDERER_MODE=remote` and `DIAGRAMS_RENDERER_URL`. `DIAGRAMS_IMAGE_STORE_DIR` can specify a temporary image directory, and `BASE_URL` can provide a public base URL for absolute download links.

## Tools and capabilities

The Diagrams MCP MCP server provides these documented operations:

- Discover providers, provider services, node classes, and nodes matching a search query.
- Render mingrammer/diagrams Python code as a cloud architecture image.
- Render Mermaid definitions as PNG or SVG.
- Render PlantUML definitions as PNG.
- Find equivalent infrastructure nodes across providers.
- List infrastructure role categories and their mapped nodes.

The listed provider and node examples include `aws`, `gcp`, `k8s`, `azure`, `onprem`, `compute`, `database`, `network`, `EC2`, and `ComputeEngine`. The exact available nodes come from the supported diagrams provider library and can be inspected through discovery tools.

## Limitations and notes

Local rendering depends on external binaries. Installing only Graphviz does not provide Mermaid or PlantUML rendering, and Mermaid or PlantUML dependencies are unnecessary if those engines are not used. The server does not itself replace the underlying diagrams, Mermaid, or PlantUML syntax; agents still need to supply valid code or definitions for the chosen renderer.

Remote mode requires a separately running renderer service and a reachable renderer URL. The repository includes Docker and Railway configuration examples for the facade and renderer roles, but deployment details depend on the hosting environment.

_Full upstream README: https://allmcps.com/mcp/diagrams-mcp/readme_