MCP Servers

模型上下文协议服务器、框架、SDK 和模板的综合目录。

Computational scrying glass — MCP server for visual computing with embedded Python, resvg rendering, and live web gallery

创建于 2/25/2026
更新于 about 4 hours ago
Repository documentation and setup instructions

scry — computational scrying glass

Computational scrying glass — an MCP server that lets Claude generate SVG visuals via Python code, render them to PNG, and serve them in a live web gallery.

Architecture

scry-mcp architecture diagram

Tools

whiteboard — Execute Python code to generate SVG visuals on a named board. Call svg('<svg>...</svg>') to set content. Variables persist between calls.

whiteboard_list — List all active boards with thumbnails, URLs, and metadata.

Requirements

  • Rust 1.85+ (edition 2024)
  • Python 3.x with development headers (PyO3 embeds CPython at build time)
  • System fonts — resvg uses fontdb to load system fonts for text rendering

Built and tested on Arch Linux (kernel 6.18, Python 3.14, Rust 1.90). Should work on other Linux distros and macOS but hasn't been tested yet — reports welcome.

Install

cargo install scry-mcp

Or build from source:

git clone https://github.com/tobert/scry-mcp
cd scry-mcp
cargo build --release

On Debian/Ubuntu you may need python3-dev for PyO3:

sudo apt install python3-dev

Add to Claude Code

claude mcp add -s user scry -- /path/to/scry-mcp --address 127.0.0.1 --port 3333

Or if installed via cargo install:

claude mcp add -s user scry -- scry-mcp --address 127.0.0.1 --port 3333

Or manually add to your MCP config:

{
  "mcpServers": {
    "scry": {
      "command": "scry-mcp",
      "args": ["--address", "127.0.0.1", "--port", "3333"]
    }
  }
}

Usage

Once configured, ask Claude to use the whiteboard tool:

"Draw a red circle on a white background using the whiteboard tool"

The gallery is live at http://localhost:3333/gallery/ — it auto-refreshes via SSE when boards update.

CLI Options

scry-mcp [OPTIONS]

Options:
      --address <ADDRESS>  Gallery bind address [default: 127.0.0.1]
      --port <PORT>        Gallery port [default: 3333]

Python Environment

Each board gets a persistent Python namespace with these pre-imported:

math, random, json, re, textwrap, itertools, functools, collections, colorsys, hashlib, string, dataclasses

WIDTH and HEIGHT are set to board dimensions (default 800x600).

Dangerous modules (os, subprocess, socket, etc.) are blocked.

Examples

All of these were generated by Claude using scry's whiteboard tool.

Data Visualization

bar chart example

Generative Art

golden ratio spirograph

Simulation

Conway's Game of Life
快速设置
此服务器的安装指南

安装命令 (包未发布)

git clone https://github.com/tobert/scry-mcp
手动安装: 请查看 README 获取详细的设置说明和所需的其他依赖项。

Cursor 配置 (mcp.json)

{ "mcpServers": { "tobert-scry-mcp": { "command": "git", "args": [ "clone", "https://github.com/tobert/scry-mcp" ] } } }