Reference¶

Tools, resources, prompts, and Python modules exposed by UML-MCP. Use the table to open one topic per page.

Section	What's here
MCP tools	`generate_uml`, `validate_uml`, `list_diagram_types`, `generate_uml_batch`. Parameters, returns, examples, fallback behavior.
MCP resources	Full `uml://` URI catalog on one page: core discovery URIs plus Mermaid- and BPMN-focused resources (see the page for names and payloads).
MCP prompts	Registered named prompts for plan-then-generate, Mermaid sequence/Gantt, BPMN, C4, WireViz, class-to-Mermaid, and more.
Python API	Auto-generated reference for `mcp_core` and `tools.kroki` modules (mkdocstrings).
Configuration	Environment variables, MCP client config, fallback knobs, server card, Smithery / Vercel options.

How the surfaces fit together¶

flowchart LR
    subgraph Client[Client]
        IDE[Cursor / Claude / curl]
    end
    subgraph Server[UML-MCP server]
        T[Tools]
        R[Resources]
        P[Prompts]
        Cfg[Config / env]
    end
    subgraph Backends[Backends]
        K[Kroki]
        PU[PlantUML]
        MM[Mermaid.ink]
    end
    IDE -->|tools/call| T
    IDE -->|resources/read| R
    IDE -->|prompts/get| P
    T -->|render| K
    K -.fallback.-> PU
    K -.fallback.-> MM
    Cfg --> T
    Cfg --> R

Tools are the side-effecting surface (rendering, validation, batching).
Resources are read-only data (types, templates, examples, server info).
Prompts are named text templates the client can fetch with prompts/get and pass into the model context.
Configuration controls runtime behavior (servers, fallback, output, rate limits).

Live discovery

tools/list, resources/list, prompts/list always reflect what's registered in the running server. The static pages below are kept in sync with the source files referenced in each section.