svgscholar

MCP server that generates publication-quality SVG diagrams, figures, plots, posters, and images for research papers.

Uses Gemini 3.1 Pro for SVG generation and Nano Banana 2 (Gemini 3.1 Flash Image) for image generation and editing.

Quick Start

Get a Gemini API key from Google AI Studio, then add the server to your client:

Claude Code

claude mcp add svgscholar -t stdio -e GEMINI_API_KEY=your-key -- npx -y svgscholar

Cursor

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "svgscholar": {
      "command": "npx",
      "args": ["-y", "svgscholar"],
      "env": {
        "GEMINI_API_KEY": "your-key"
      }
    }
  }
}

Windsurf / VS Code / Other MCP Clients

Add to your MCP config file:

{
  "mcpServers": {
    "svgscholar": {
      "command": "npx",
      "args": ["-y", "svgscholar"],
      "env": {
        "GEMINI_API_KEY": "your-key"
      }
    }
  }
}

Vertex AI (Alternative)

Use Vertex AI instead of an API key:

claude mcp add svgscholar -t stdio \
  -e GOOGLE_CLOUD_PROJECT=your-project \
  -e GOOGLE_CLOUD_LOCATION=us-central1 \
  -- npx -y svgscholar

Tools

generate_svg

Generate publication-ready SVG visuals: architecture diagrams, conference posters, paper figures, data plots, LaTeX equations, presentation slides, and teaser/hero figures.

Parameters: | Parameter | Required | Description | |-----------|----------|-------------| | type | Yes | diagram · poster · figure · plot · equation · slides · teaser | | description | Yes | Detailed description of what to generate | | options | No | Type-specific options (e.g., data for plots, latex for equations) | | filename | No | Output filename (default: auto-generated) |

generate_image

Generate raster images (PNG) for concept illustrations, visual abstracts, or photorealistic content.

Parameters: | Parameter | Required | Description | |-----------|----------|-------------| | description | Yes | What to generate | | filename | No | Output filename |

edit_image

Annotate existing images with arrows, labels, bounding boxes, or other overlays.

Parameters: | Parameter | Required | Description | |-----------|----------|-------------| | image_path | Yes | Absolute path to the input image | | edit_description | Yes | What edits to make | | filename | No | Output filename |

All outputs are saved to a svgscholar-output/ directory in your current working directory.

Example Prompts

• "Draw my model architecture — we have a shared encoder that feeds into two task-specific heads with a gating mechanism between them"
• "I need a figure comparing our method against three baselines across four datasets, like a grouped bar chart with error bars"
• "Make a poster for my EMNLP submission on cross-lingual transfer, here's the abstract and key results..."
• "Can you turn this pipeline into a proper figure: we take raw clinical notes, run NER, link to UMLS, then feed into a graph neural network"
• "Plot our ablation results from this table and make it camera-ready"
• "I need a teaser figure that shows the gap between supervised and zero-shot, and how our method closes it"
• "Render the ELBO objective with the KL term expanded"
• "Annotate this attention map — add arrows pointing to the key tokens and label the heads"

Configuration

| Variable | Required | Description | |----------|----------|-------------| | GEMINI_API_KEY | Yes* | Gemini API key from AI Studio | | GEMINI_MODEL | No | Model for SVG generation (default: gemini-3.1-pro-preview) | | GEMINI_VISION_MODEL | No | Model for image generation/editing — Nano Banana 2 (default: gemini-3.1-flash-image-preview) | | GOOGLE_CLOUD_PROJECT | No | GCP project ID (for Vertex AI) | | GOOGLE_CLOUD_LOCATION | No | GCP region (for Vertex AI, default: us-central1) |

*Required unless using Vertex AI.

Development

bun install
bun run src/index.ts

License

MIT