Data & analysis

diagram-generator

Try it

Generate and edit Draw.io, Mermaid, and Excalidraw diagrams from natural language using a structured JSON spec.

What it does

Translates a user's diagram request into a JSON specification and passes it to the mcp-diagram-generator MCP server, which writes the actual file. Supports new diagrams and edits to existing .drawio, .mmd, and .excalidraw files, with default output paths under diagrams/{format}/. Includes format-specific playbooks for network topology, architecture, flowcharts, swimlanes, UML sequence/class/ER, and whiteboard sketches, plus a quality gate that checks format, geometry, container hierarchy, and edge placement before generation.

When to use it

  • Drafting a network topology diagram for a datacenter zone layout
  • Editing an existing .drawio file to add new nodes and edges
  • Generating a Mermaid sequence or ER diagram for a code repository
  • Producing a hand-drawn style Excalidraw whiteboard sketch

The skill document

Diagram Generator

Purpose

Create and edit diagrams by converting user intent into a structured JSON specification, then delegating file generation to the mcp-diagram-generator MCP server.

Supported formats:

  • Draw.io: .drawio
  • Mermaid: .mmd or markdown Mermaid content
  • Excalidraw: .excalidraw

Supported work:

  • Natural-language diagram creation
  • Existing .drawio, .mmd, and .excalidraw edits
  • Default output paths under diagrams/{format}/
  • Custom filenames and output paths

Required MCP Tools

Before generating a diagram, verify that the MCP server tools are available:

  • mcp__mcp-diagram-generator__get_config
  • mcp__mcp-diagram-generator__generate_diagram
  • mcp__mcp-diagram-generator__init_config

If the tools are missing, configure the MCP server.

Recommended remote configuration:

{
  "mcpServers": {
    "mcp-diagram-generator": {
      "command": "npx",
      "args": ["-y", "mcp-diagram-generator"]
    }
  }
}

After changing configuration, restart the agent environment. On first use, the server creates .diagram-config.json and default output directories. For local development configuration, see the "Tool missing" entry in Troubleshooting.

Main Workflow

1. Intake

For new diagrams, collect the basic options before accepting or processing the full diagram prompt:

  • Diagram type
  • Output format
  • Layout direction
  • Usage context
  • Optional filename or output directory

Read references/interaction-intake-guide.md before asking intake questions.

Skip intake when the user already provided all required options and the full prompt. For existing-file edits, ask only for the target file and requested changes if missing.

2. Dispatch To A Playbook

Select exactly one primary playbook based on the diagram type:

User IntentPrimary Playbook
Network topology, datacenter, zone, router, switch, firewallreferences/playbook-network-topology.md
System architecture, application architecture, layered component diagramreferences/playbook-architecture.md
Flowchart, process, decision treereferences/playbook-flowchart.md
Swimlane, cross-team handoff, approval workflow by departmentreferences/playbook-swimlane.md
Sequence, class, ER, UML-style diagramsreferences/playbook-uml.md
Whiteboard sketch, hand-drawn style, informal Excalidraw diagramreferences/playbook-excalidraw.md
Unsure about formatreferences/format-selection-guide.md first, then the matching playbook

Only read the playbook needed for the current diagram. If a playbook points to json-schema-guide.md or network-topology-examples.md, read only the relevant section. For explicit geometry, also read references/layout-quality-guide.md.

3. Choose Format

Use these defaults unless the user explicitly chooses otherwise:

Diagram TypeDefault FormatDefault Direction
Network topologyDraw.ioVertical
ArchitectureDraw.ioVertical or automatic
FlowchartMermaidVertical
SwimlaneDraw.ioHorizontal
SequenceMermaidAutomatic
ClassMermaidAutomatic
ERMermaidAutomatic
Whiteboard sketchExcalidrawAutomatic

Usage context can override defaults:

  • Word: prefer portrait-friendly vertical layouts.
  • PPT: horizontal layouts are acceptable when readability improves.
  • Code repositories and documentation: prefer Mermaid for simple flow, sequence, class, and ER diagrams.
  • Whiteboard collaboration: prefer Excalidraw.
  • Complex network or architecture diagrams: prefer Draw.io unless the user explicitly asks for Excalidraw.

4. Build The JSON Specification

Follow references/json-schema-guide.md for the schema. Core structure:

{
  "format": "drawio",
  "diagramType": "architecture",
  "title": "Diagram title",
  "elements": [
    {
      "id": "unique-id",
      "type": "container",
      "name": "Display name",
      "level": "environment",
      "geometry": { "x": 0, "y": 0, "width": 800, "height": 600 },
      "children": []
    },
    {
      "type": "edge",
      "source": "source-id",
      "target": "target-id"
    }
  ]
}

Universal rules:

  • elements must be an array.
  • IDs must be unique.
  • Edges must be top-level elements, never inside children.

Full field reference: references/json-schema-guide.md.

5. Quality Gate

Before calling the MCP server, verify:

  • The chosen format matches the intake answer and playbook.
  • diagramType is explicit when supported.
  • Layout direction is reflected in coordinates or generator-specific fields.
  • Complex Draw.io and Excalidraw diagrams have explicit geometry.
  • Container hierarchy is valid.
  • Edges are top-level elements.
  • Text and connector rules for the selected format are followed.

After generation, inspect the saved file enough to confirm the expected format-specific properties exist. For code changes to the MCP server, also run npm run test:diagrams from mcp-diagram-generator/.

6. Generate

Preferred call:

{
  "diagram_spec": ""
}

Optional filename:

{
  "diagram_spec": "",
  "filename": "my-diagram.drawio"
}

Optional explicit output path:

{
  "diagram_spec": "",
  "output_path": "custom/path/to/diagram.drawio"
}

The MCP server validates the schema, creates missing directories, and writes to the configured default directory when no output path is supplied.

Configuration Helpers

Initialize defaults:

init_config()

Set custom paths:

{
  "paths": {
    "drawio": "output/diagrams/drawio",
    "mermaid": "output/diagrams/mermaid",
    "excalidraw": "output/diagrams/excalidraw"
  }
}

Inspect configuration:

get_config()

Update one format path:

{
  "format": "drawio",
  "path": "custom/drawio-path"
}

Troubleshooting

Tool missing:

  • Configure the MCP server with the remote (npx) block above and restart the agent environment.
  • For local development, use "command": "node", "args": ["/absolute/path/to/mcp-diagram-generator/dist/index.js"] instead.

Schema validation failed:

  • Read references/json-schema-guide.md.
  • Check required fields, unique IDs, edge source/target, and parent-child structure.

Directory error:

  • Check write permissions.
  • Run get_config().
  • Reinitialize with init_config() if needed.

Wrong extension:

  • Draw.io uses .drawio.
  • Mermaid uses .mmd or markdown output.
  • Excalidraw uses .excalidraw.

Nested container issue:

  • Child coordinates are relative to the direct parent.
  • Container sizes must fit child bounds plus padding.
  • Network topology must follow environment -> datacenter -> zone -> device.

Reference Index

Read only what is needed:

  • references/interaction-intake-guide.md: interactive intake defaults and question template.
  • references/format-selection-guide.md: format selection matrix.
  • references/playbook-network-topology.md: Draw.io and Excalidraw network topology rules.
  • references/playbook-architecture.md: layered architecture rules.
  • references/playbook-flowchart.md: process and decision flow rules.
  • references/playbook-swimlane.md: swimlane and handoff rules.
  • references/playbook-uml.md: sequence, class, and ER rules.
  • references/playbook-excalidraw.md: Excalidraw whiteboard and binding rules.
  • references/layout-quality-guide.md: explicit geometry, spacing, and connector readability rules.
  • references/json-schema-guide.md: schema details and examples.
  • references/network-topology-examples.md: network topology JSON patterns.

Output Discipline

When responding to the user:

  • Confirm the selected diagram type, format, direction, and output file.
  • Do not paste the full JSON unless the user asks.
  • Provide the saved file path.
  • Mention any validation or regression command that was run.

Questions people ask

Which diagram formats are supported?
Draw.io (.drawio), Mermaid (.mmd or markdown Mermaid content), and Excalidraw (.excalidraw), each with a default output directory under diagrams/{format}/.
Does it require the mcp-diagram-generator MCP server?
Yes. The skill calls mcp__mcp-diagram-generator__get_config, __generate_diagram, and __init_config, so the server must be configured before use.
Can it edit existing diagram files?
Yes. Pass the target .drawio, .mmd, or .excalidraw file plus the requested changes, and the server produces an updated file at the same or a specified path.

Related skills

Excalidraw MCP (excalidraw.com). Use this skill for ANY Excalidraw MCP request — reading, creating, and updating data. Whenever a task involves Excalidraw MCP, use this skill instead of calling the API directly.

1 installs

Generate beautiful draw.io diagrams with a mandatory visual QA loop. Covers flowcharts, ERD, architecture, sequence, and class diagrams. Use when user reques...

Create and edit draw.io diagrams through the configured drawio MCP server, including flowcharts, architecture diagrams, ML model diagrams, Chinese labels, an...

54 installs

Generate beautiful draw.io diagrams with a mandatory visual QA loop. Covers flowcharts, ERD, architecture, sequence, and class diagrams. Use when user reques...

by Spotlight Revenue

AI diagram and flowchart generation powered by CellCog. Flowcharts, system architecture, mind maps, org charts, ER diagrams, sequence diagrams, Gantt charts, network diagrams. Describe your system in plain English, get interactive diagrams or print-ready PDFs.

22 installs