
heroicons-mcp
heroicons-mcp
README
heroicons-mcp
A Model Context Protocol (MCP) server exposing Heroicons as resources and tools for LLMs and agentic applications. Built with Bun and the MCP TypeScript SDK.
What is Heroicons?
Heroicons is a popular library of hand-crafted SVG icons, designed by the makers of Tailwind CSS. The icons are available in multiple styles (Outline, Solid) and are easy to integrate into web projects.
What is MCP?
The Model Context Protocol (MCP) is a standard for AI tools to request specific context from sources outside their main training data.
This MCP server allows AI coding assistants and other agentic applications to access information about Heroicons, enabling better assistance and icon search capabilities.
Features
- Exposes Heroicons as MCP resources (Outline and Solid styles)
- Provides tools for searching icons by name or keywords
- Allows listing all icons or icons within a specific style
- Ready for integration with Claude Desktop and other MCP clients
- Can be run as an HTTP server or a stdio-based MCP server
Prerequisites
Getting Started (Development)
1. Clone the repository
git clone https://github.com/SeeYangZhi/heroicons-mcp.git
cd heroicons-mcp
2. Install Bun (if you don't have it)
Refer to the official Bun installation guide.
After installation, restart your terminal and check:
bun --version
3. Install dependencies
bun install
4. Build the project
This compiles the TypeScript source to JavaScript in the build
directory.
bun run build
Usage
HTTP Mode
You can run the HTTP server using npx
:
npx heroicons-mcp
This starts the HTTP server (defaults to port 3000, as defined in src/http.ts
).
Or install globally:
npm install -g heroicons-mcp
Then run:
heroicons-mcp
Stdio Mode
npx heroicons-mcp --stdio
# or if installed globally
heroicons-mcp --stdio
Local Development
There are two main ways to run the MCP server:
1. HTTP Mode
Suitable for clients that support communication over HTTP.
For development (using Bun):
bun run start
# or directly
bun run src/entry.ts
This runs the server defined in src/entry.ts
, which defaults to HTTP mode.
2. Stdio Mode
Often used for direct integration with tools like Claude Desktop or the MCP Inspector, communicating over standard input/output.
For development (using Bun):
bun run src/entry.ts --stdio
Configuration with AI Tools
Example: Claude Desktop
To use this MCP server in Claude Desktop:
- Open your Claude Desktop configuration file:
code ~/Library/Application\ Support/Claude/claude_desktop_config.json
(Or use your preferred editor) 2. Add the server to the mcpServers
section.
Option A: via npx
:
{
"mcpServers": {
"heroicons": {
"command": "npx",
"args": ["heroicons-mcp", "--stdio"]
}
}
}
Option B: Pointing directly to the build output (ensure you have built the project using bun run build
):
{
"mcpServers": {
"heroicons": {
"command": "node",
"args": ["/ABSOLUTE/PATH/TO/heroicons-mcp/build/entry.js", "--stdio"]
}
}
}
Replace /ABSOLUTE/PATH/TO/heroicons-mcp/build/entry.js
with the actual absolute path to your built entry.js
file.
- Save the file and restart Claude Desktop.
- You should now see the "heroicons" server available in Claude's tools panel.
Note: The npx heroicons-mcp --stdio
command is the recommended way for stdio mode.
Tools Available (MCP)
This MCP server exposes the following tools to AI coding assistants:
- list_all_icons
- Description: Lists all available Heroicons, optionally filtered by style (outline, solid).
- Parameters:
style
(optional: "outline" | "solid")
- search_icons
- Description: Searches for Heroicons by name or keywords across all styles.
- Parameters:
query
(string),style
(optional: "outline" | "solid")
- get_icon_usage_examples
- Description: Retrieves JSX example usage for a specific icon.
- Parameters:
name
(string),style
(string: "outline" | "solid")
Example Usage
Here's how an AI tool might use this MCP server:
- User asks AI tool: "Find me a 'user' icon from Heroicons, preferably the solid style."
- AI tool calls
search_icons
:
query
: "user"style
: "solid"
- MCP server responds with a list of matching solid Heroicons (e.g.,
UserIcon
,UserCircleIcon
,UserPlusIcon
). - User asks tool: "Show usage example of UserIcon".
- AI tool calls
get_icon_usage_examples
:
name
: "UserIcon"style
: "solid"
- MCP server responds with the JSX code example:
import { UserIcon } from "@heroicons/react/24/solid";
function Example() {
return (
<div>
<UserIcon className="w-6 h-6 text-blue-500" />
</div>
);
}
Testing MCP Locally with Inspector
You can test the MCP server (stdio mode) locally using the MCP Inspector.
First, ensure the project is built:
bun run build
Then launch the Inspector and connect it to your server using the node ./build/entry.js
command with the --stdio
flag:
npx @modelcontextprotocol/inspector node ./build/entry.js --stdio
This will open the Inspector interface, allowing you to interactively test resources and tools exposed by your MCP server.
Development Scripts
bun run dev
: Starts the server in HTTP mode for development (usessrc/entry.ts
).bun run dev:stdio
: Starts the stdio MCP server for development (usessrc/entry.ts --stdio
).bun run build
: Compiles TypeScript to JavaScript (output inbuild/
).bun run lint
: Lints the codebase using ESLint.
Resources
License
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。