That uniformity is the useful part — the input format stops being a decision about which API to learn and becomes an argument. There is no editing, layout or re-render tool here, so a diagram that comes back wrong is revised in the browser page the call returns rather than through a second call. generate_diagram_from_mermaid is the odd member: it changes notation rather than medium, taking Mermaid source to D2 and returning a PNG.
A diagram generator split by input format — natural-language description, JSON structure, ASCII art, image or Mermaid source — returning a link to view and edit the generated diagram in the browser.
- generate_diagram_from_text draws from a natural-language description of code, systems, processes or data flows, with diagramType selecting the form: flowchart, sequence, ERD, system architecture and network architecture are among those supported.
- generate_diagram_from_json takes a raw JSON string as content and renders structured data directly — API responses, database schemas, dependency trees, configuration files.
- generate_diagram_from_ascii accepts box-drawing characters, arrow forms such as --> and ==>, and plain text layouts, and returns them as a drawn diagram.
- generate_diagram_from_image converts a whiteboard photo, screenshot or hand-drawn sketch, supplied as a public image URL or a base64 data URI.
- generate_diagram_from_mermaid converts existing Mermaid source into a D2 diagram and returns a PNG image.
- Every call takes the same arguments whatever the input format — content, diagramType, prompt and isIconEnabled — so switching source material does not mean rewriting the call.
An AI Diagram Maker API key, supplied as ADM_API_KEY. Images have to arrive as a public URL or an inline base64 data URI, so a local screenshot needs to be encoded or hosted first.
One command plus a key — ADM_API_KEY=your_api_key npx ai-diagram-maker-mcp@latest, then supply credentials