
Mcp Icon Visual
io.github.n0isy由 GitHub 账号 n0isy 发布v0.1.1更新于 Oct 24, 2026
Search, retrieve, compare, and render SVG icons visually for AI agents.
概览
一个本地 MCP 服务器,让具备视觉能力的助手搜索、获取、比较 SVG 图标,并以 PNG 网格形式可视化渲染。
- 可以做什么
- 按关键词搜索图标,返回 4x4 PNG 网格,并给出单元格编号到 SVG 链接的对应关系,让助手凭视觉挑选而非盲选。从 URL、Iconify ID、本地文件路径或内联 SVG 字符串获取原始 SVG 标记。将 1 至 16 个 SVG 来源渲染到 512x512 的对比网格上,背景色可配置,用于并排比较风格或迭代修改。
- 适用场景
- 适合具备视觉能力的助手为界面挑选图标、让新图标匹配项目既有视觉风格,或在多轮迭代中对比 SVG 修改效果。由于工具返回的是图片,对纯文本模型用处不大。
- 运行要求
- 以 stdio 在本地运行,通常通过 npx 使用 npm 包,或使用 Rust 编译的二进制文件。需要访问远程图标渲染 API 的网络连接(默认上游服务,或通过 api-base 参数指定自定义端点)。未声明需要账号、API 密钥或环境变量。
安装
在 SourceWeft 中
- 打开 控制台中的 Mcp Icon Visual,将其添加到工作区。
- 为需要使用其工具的对话启用该服务。
Desktop only,通过 STDIO。 STDIO 服务会启动本地进程,因此需要 SourceWeft 桌面宿主。
其他 MCP 客户端
参照 仓库 中的启动说明。
README
mcp-icon-visual
[MCP Registry] [npm] [License: MIT]
MCP server that gives AI agents the ability to search, retrieve, compare, and render SVG icons. Built with Rust and the rmcp SDK.
Combines local filesystem access with a remote rendering API to resolve SVGs from any source — inline strings, URLs, local files, or Iconify IDs — and render them onto visual comparison grids.
The key insight: every tool returns images. A vision-capable LLM doesn't just get metadata — it sees the icons. This turns icon selection from guesswork into a visual conversation.
[search_icons result for 'machine learning' — 4x4 grid with 16 icons]
search_icons("machine learning") — 4x4 grid, cells 0–15, returned as PNG
Agent Pipelines
These tools are designed to be composed by AI agents with vision. Below are three proven patterns.
1. Visual Search & Selection
An agent searches for icons and visually picks the best match — no blind URL selection.
The agent runs multiple searches, looks at each grid, picks candidates by cell number, then renders a final side-by-side comparison to make the choice. Works especially well when the agent has context about the project's visual style.
2. Iterative Icon Editing
An agent modifies SVG code and uses render_grid as a visual diff to track progress across iterations.
Each iteration, the agent passes all versions (original + edits) as inline SVGs to render_grid. It sees them side-by-side in one image and decides whether to keep iterating or stop. The original always stays in cell 0 as a reference.
3. Style Matching & Consistency
An agent ensures new icons match the visual style of existing project icons.
By mixing local file paths and remote sources in a single grid, the agent can visually compare existing project icons against candidates in one shot. The grid becomes a style audit tool.
Combining Pipelines
These patterns compose naturally. A real-world workflow might:
- Search across multiple keywords (pipeline 1)
- Compare top candidates against existing project icons (pipeline 3)
- Edit the closest match to fix style inconsistencies (pipeline 2)
- Verify the final icon one more time against the full icon set (pipeline 3)
All driven by a single agent with vision, using three tools.
Tools
search_icons
Semantic icon search. Returns a 4x4 PNG grid (512x512, cells 0-15) and a mapping of cell numbers to SVG URLs.
get_svg
Retrieve raw SVG markup from any source. Returns the complete SVG string for inspection or modification.
render_grid
Render 1-16 SVGs onto a 4x4 comparison grid (512x512, cells 0-15). Useful for comparing icon variants side-by-side or previewing local icons alongside search results.
Source Resolution
All tools share a unified resolution pipeline that classifies sources automatically:
File paths are detected by prefix (/, ./, ~/) or .svg extension. Everything else without http:///https:// is treated as an Iconify ID.
Grid Format
All grid outputs are 512x512 PNG images with a 4x4 layout (128x128 cells), numbered 0-15 left-to-right, top-to-bottom:
Cell numbers appear at the bottom-left of each cell. Unused cells show the number only.
Installation
npx (recommended)
No build needed. Works on Linux (x64/arm64), macOS (Intel/Apple Silicon), and Windows (x64).
From source
Requires Rust (edition 2024):
The release profile produces a fully static binary (musl libc, rustls-tls, LTO, stripped).
Usage
The server communicates over stdio using the MCP protocol.
MCP Client Configuration
Claude Desktop / Cursor
Custom API endpoint
Project Structure
Testing
Dependencies
License
MIT License
Copyright (c) 2025
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
来源:README.md,提交 92d593c
工具
0版本历史
1- v0.1.1最新Sep 16, 2026
评分
已满足适用检查项的 40%
该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。
- 近 30 天内可连通25 分
请保持端点能够响应 MCP 握手;SourceWeft 每 30 天探测一次。
- 已声明认证方式10 分
请在注册中心条目中声明凭据请求头,或保持端点能够响应探测,以便判定其认证方式。
- 已列出工具15 分
请让服务器在无凭据时响应 tools/list,或发布可在沙箱中无凭据启动的软件包。
- 描述必需15 分
- 图标10 分
- 文档15 分
- 许可证10 分
请在仓库或注册中心条目中声明许可证。