Quiver-MCP
Generate SVGs from text prompts and vectorize raster images using QuiverAI's Arrow model (#1 on SVG Arena). Includes baked-in prompt engineering guidance, reference image support, and an outputPath parameter to save SVGs directly to disk. TypeScript, MIT-licensed, requires a free QuiverAI API key.
README
<p align="center"> <img src="examples/logo.svg" width="120" alt="quiver-mcp logo" /> </p>
<h1 align="center">quiver-mcp</h1>
<p align="center"> <a href="https://www.npmjs.com/package/@syntropic/quiver-mcp"><img src="https://img.shields.io/npm/v/@syntropic/quiver-mcp.svg" alt="npm version" /></a> <a href="https://www.npmjs.com/package/@syntropic/quiver-mcp"><img src="https://img.shields.io/npm/dw/@syntropic/quiver-mcp.svg" alt="npm downloads" /></a> <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-blue.svg" alt="License: MIT" /></a> </p>
<p align="center"> MCP server for <a href="https://quiverai.com">QuiverAI</a> — generate SVGs from text prompts and vectorize raster images using AI, directly from Claude (or any MCP-compatible client). </p>
Examples
Generated by Claude calling this MCP. Each took ~60s at n: 3, temperature: 0.9. Both prompts are documented in the tool description, so Claude knows the recipe.
<table> <tr> <td width="50%" align="center"> <img src="examples/fountain-pen-detailed.svg" width="280" alt="Exploded isometric fountain pen" /> </td> <td width="50%" align="center"> <img src="examples/japanese-crane-detailed.svg" width="280" alt="Japanese crane in woodblock style" /> </td> </tr> <tr> <td valign="top">
Prompt: exploded isometric view of a Montblanc Meisterstück fountain pen, technical blueprint drawing, thin line art, dotted grid background, labeled components, engineering illustration
</td> <td valign="top">
Prompt: Japanese crane in traditional woodblock illustration style with warm earth tones
Instructions: Use a warm muted palette with detailed feather work
</td> </tr> </table>
More variants in examples/.
Requirements
- Node.js 18+
- A QuiverAI API key
Installation
Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"quiverai": {
"command": "npx",
"args": ["-y", "@syntropic/quiver-mcp"],
"env": {
"QUIVERAI_API_KEY": "your_api_key_here"
}
}
}
}
Manual
npm install -g @syntropic/quiver-mcp
QUIVERAI_API_KEY=your_api_key_here quiver-mcp
Tools
generate_svg
Generate one or more SVGs from a text prompt.
| Parameter | Type | Required | Description |
|---|---|---|---|
prompt |
string | yes | Text description of the SVG to generate |
model |
string | yes | Model ID (use list_models to discover options) |
instructions |
string | no | Additional style or formatting guidance |
n |
number | no | Number of SVGs to generate (default: 1) |
temperature |
number | no | Sampling temperature 0–2 (default: 1) |
references |
array | no | Up to 4 image references ({url} or {base64}) for palette and composition guidance. Style keywords must still be in the prompt text. |
outputPath |
string | no | Absolute file path to save SVG(s) to disk. For multiple variants (n > 1), files are saved with _1, _2 … suffixes. Parent directories are created automatically. |
Prompt tips
The tool description includes extensive prompt guidance, but in short:
- Structure prompts with three parts: subject (concrete object), style (aesthetic keywords like
line art,isometric,flat monochrome), and color palette (hex codes where possible). - Use famous physical objects the model knows. Avoid abstract software concepts (
AI agent,workflow) — use physical metaphors instead. - For exploration, generate 3+ variants at
temperature: 0.9. Some generations produce corrupted tails; extra variants give you options.
vectorize_svg
Convert a raster image (PNG, JPG, etc.) to SVG.
| Parameter | Type | Required | Description |
|---|---|---|---|
model |
string | yes | Model ID |
image |
object | yes | Image to vectorize — {url} or {base64} |
autoCrop |
boolean | no | Crop to dominant subject before vectorizing (default: false) |
targetSize |
number | no | Square resize target in pixels before vectorizing |
temperature |
number | no | Sampling temperature 0–2 (default: 1) |
outputPath |
string | no | Absolute file path to save the SVG to disk. Parent directories are created automatically. |
list_models
List all available QuiverAI models with supported operations and pricing.
Environment Variables
| Variable | Description |
|---|---|
QUIVERAI_API_KEY |
Required. Your QuiverAI API key |
Development
npm install
npm run build # compile TypeScript
npm run dev # watch mode
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。