sean-perkins/mcp-swimlanes-server

MCP Swimlanes Server

MCP server that turns natural-language prompts into swimlanes.io diagrams and images.

Install

pnpm i || npm i
pnpm run build || npm run build
# Optional for local CLI use without absolute paths:
npm link

Optional: copy .env.example to .env and set at least one LLM key if you want prompt-to-syntax conversion handled server-side.

Run (standalone)

mcp-swimlanes

This server is meant to be launched by an MCP-capable client over stdio.

Tools

• swimlanes.text_from_prompt(prompt, provider?, model?) → returns Swimlanes syntax
• swimlanes.generate_link(text) → returns editor/share link
• swimlanes.generate_image_link(text, high_resolution?) → returns PNG link
• swimlanes.generate_image(text, high_resolution?) → returns base64 PNG as an image content item
• swimlanes.image_from_prompt(prompt, high_resolution?, provider?, model?) → base64 PNG
• swimlanes.link_from_prompt(prompt, provider?, model?) → editor/share link

Configuration

Environment variables (use any one provider):

• ANTHROPIC_API_KEY and optional ANTHROPIC_MODEL (default claude-3-5-sonnet-latest)
• OPENAI_API_KEY and optional OPENAI_MODEL (default gpt-4o-mini)

Example client config

Claude Desktop (macOS) claude_desktop_config.json snippet:

{
  "mcpServers": {
    "swimlanes": {
      "command": "mcp-swimlanes",
      "env": {
        "OPENAI_API_KEY": "...",
        "OPENAI_MODEL": "gpt-4o-mini",
        "ANTHROPIC_API_KEY": "...",
        "ANTHROPIC_MODEL": "claude-3-5-sonnet-latest"
      }
    }
  }
}

If you prefer not to use the CLI binary, you can still point directly to the built entry:

{
  "mcpServers": {
    "swimlanes": {
      "command": "node",
      "args": ["/Users/user/absolute/path/to/mcp-swimlanes/dist/index.js"],
      "env": {
        "OPENAI_API_KEY": "..."
      }
    }
  }
}

Cursor supports MCP servers via "Tools". Add a tool with command mcp-swimlanes (recommended) or node + args dist/index.js, and set env vars.

Notes

• The server reads the local syntax.md and provides it to the LLM to ensure valid output.
• For best print quality, set high_resolution: true.