Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@RemixIcon MCPfind icons for user, profile, and account"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
RemixIcon MCP
English | 简体中文
A lightweight Model Context Protocol (MCP) server that maps icon-focused keywords directly to Remix Icon metadata. Provide concise keywords (up to 20), receive the top 5 matching icon names and metadata – clean architecture with FlexSearch-powered local search.
Features
Smart Keyword Input – Supports up to 20 comma-separated keywords while rejecting natural-language sentences for optimal search quality.
Fixed Top-5 Results – Returns exactly 5 most relevant icons for focused decision-making.
FlexSearch-backed Index – Uses FlexSearch v0.8's document index for high-performance token lookup over the local Remix Icon catalog.
Clean Architecture – Domain entities, application use cases, infrastructure adapters, and MCP interface remain isolated for easy testing.
CLI Ready – Can be run as a standalone CLI tool via
npx remixicon-mcpor integrated into MCP clients.LLM-ready Responses – Returns ranked candidates, matched tokens, and explicit guidance instructing the model to choose exactly one icon.
Quick Start
Installation
Usage
As a Standalone CLI Tool
You can run the MCP server directly via stdio for testing or integration:
Platform Setup
Claude Desktop
Configuration
Add the following to your claude_desktop_config.json:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Setup Steps
Save the configuration file
Completely quit and restart Claude Desktop
The
search_iconstool will be available in your conversations
Claude Code
Option 1: Marketplace Plugin (Recommended)
Benefits:
Automatic installation and updates
Complete plugin metadata and versioning
Rich discovery with keywords and categorization
Full integration with Claude Code's plugin ecosystem
Option 2: Manual Configuration
Or manually add to your project's .claude/settings.json:
Setup Steps
Choose one of the installation methods above
Restart Claude Code for the changes to take effect
The
search_iconstool will be available in your sessions
Codex
Configuration
Setup Steps
Run the installation command above
Restart Codex for the changes to take effect
The
search_iconstool will be available in your conversations
Available Tools
The server communicates over stdio using JSON-RPC 2.0 via the official @modelcontextprotocol/sdk and exposes a single tool:
search_icons
Input: keywords string (comma-separated, up to 20 keywords)
Output: Top 5 most relevant icons with metadata
Format: Human-readable summary + structured metadata
Example Usage
JSON-RPC Call:
Sample Response: The server returns the top 5 icons that match your keywords, complete with names, categories, and usage information.
Project Structure
Implementation Notes
Keywords are parsed with Unicode-aware boundaries, supporting up to 20 comma-separated keywords while rejecting sentence-style inputs.
Enhanced detection differentiates between keyword lists (with delimiters) and natural language sentences (space-separated phrases).
FlexSearch indexes icon names, tags, usage, and categories; field weights plus token matches drive deterministic scores.
Fixed top-5 results provide focused, relevant matches without configuration complexity.
The application layer combines parser validation, repository queries, and response formatting so the interface only handles transport concerns.
MCP responses include natural-language guidance and machine-readable matches so LLM clients can choose exactly one icon.
CLI runner enables standalone execution via
npxor global installation for easy integration.