
kenneth-liao/mcp-launchpad 是从 https://github.com/kenneth-liao/mcp-launchpad. 导入的 MCP 服务器候选
概览
kenneth-liao/mcp-launchpad 是从 https://github.com/kenneth-liao/mcp-launchpad. 导入的 MCP 服务器候选
概述(AI 翻译)
一个轻量级CLI用于发现和执行来自多个MCP(模型上下文协议)服务器的工具。
安装
前提条件:uv - Python 包和环境管理器
验证它有效:
快速参考
配置
在你的项目目录中创建mcp.json(全局配置时创建[~/.claude/mcp.json):
用mcpl list进行验证。如果看不到服务器,重启终端并运行mcpl list --refresh。
增加更多服务器
环境变量使用 ${VAR} 语法,加载自 ~/.claude/.env 和 ./.env。
配置发现
MCP Launchpad 按以下顺序搜索配置文件:
./mcp.json- 项目级配置./.claude/mcp.json- 项目级 Claude 配置~/.claude/mcp.json- 用户级配置
当发现多个配置文件时,mcpl会提示你选择使用哪些配置文件。你的偏好设置会被保存,之后可以用mcpl config files --select更改。
代理集成
MCP Launchpad 支持任何能运行 bash 命令的 AI 代理(Claude Code、Cursor、Windsurf 等)。
将根 CLAUDE.md 内容复制到你的 CLAUDE.md 或 AGENTS.md] 中,以教你的代理如何使用 mcpl。
本页由 AI 翻译,技术细节请以原文为准。
README
以下为原文 README 节选,完整内容请查看原文。
MCP Launchpad
A lightweight CLI for discovering and executing tools from multiple MCP (Model Context Protocol) servers.
Installation
Prerequisite: uv - Python package and environment manager
uv tool install https://github.com/kenneth-liao/mcp-launchpad.git
Verify it works:
mcpl --help
Quick Reference
# Find tools
mcpl search "" # Search all tools (returns 5 by default)
mcpl search "" --limit 10 # Get more results
mcpl list # List all MCP servers
mcpl list # List tools for a server
# Get tool details
mcpl inspect # Full schema
mcpl inspect --example # Schema + example call
# Execute tools
mcpl call '{}' # No arguments
mcpl call '{"param": "value"}' # With arguments
# Troubleshooting
mcpl verify # Test all server connections
mcpl session status # Check daemon status
mcpl session stop # Restart daemon (auto-restarts on next call)
Configuration
Create mcp.json in your project directory (or ~/.claude/mcp.json for global config):
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@anthropic/mcp-server-filesystem", "/path/to/allowed/dir"]
}
}
}
Validate with mcpl list. If you don’t see your servers, restart your terminal and run mcpl list --refresh.
Adding More Servers
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@anthropic/mcp-server-filesystem", "/path/to/dir"]
},
"github": {
"command": "uvx",
"args": ["mcp-server-github"],
"env": {
"GITHUB_TOKEN": "${GITHUB_TOKEN}"
}
}
}
}
Environment variables use ${VAR} syntax and are loaded from ~/.claude/.env and ./.env.
Config Discovery
MCP Launchpad searches for config files in this order:
./mcp.json- Project-level config./.claude/mcp.json- Project-level Claude config~/.claude/mcp.json- User-level config
When multiple config files are found, mcpl prompts you to select which ones to use. Your preferences are saved and can be changed later with mcpl config files --select.
Agent Integration
MCP Launchpad works with any AI agent that can run bash commands (Claude Code, Cursor, Windsurf, etc.).
Copy the root CLAUDE.md contents into your own CLAUDE.md or AGENTS.md to teach your agent how to use mcpl. This can be done at the project level (./CLAUDE.md) or user level (~/.claude/CLAUDE.md).
Important: OAuth Authentication
安装
npx -y @anthropic/mcp-server-filesystem /path/to/allowed/dir配置
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@anthropic/mcp-server-filesystem", "/path/to/allowed/dir"]
}
}
}