MCP Opener
Enables opening files, folders, and URLs across different operating systems with special support for Linux Flatpak Firefox installations, handling filesystem sandboxing restrictions automatically.
README
MCP Opener
Model Context Protocol server for opening files, folders, and browsers across different operating systems. Designed specifically to handle Linux Flatpak Firefox complexities on distributions like Fedora Bluefin.
Features
- Open Folders: Launch system file manager (Nautilus, Finder, Explorer, etc.)
- Open Files: Open files with their default applications
- Open Browser: Open URLs with configurable browser preferences
- Flatpak Firefox Support: Auto-detects and handles Flatpak Firefox installations
- Cross-Platform: Works on Linux, macOS, and Windows
- Smart File Handling: Automatically copies files to
/tmpfor Flatpak browser access when needed
Installation
# Install dependencies
pnpm install
# Build
pnpm build
# Install globally (optional)
pnpm link --global
Configuration
Configure via environment variables in your .env file or Claude Code MCP settings:
# Preferred browser: 'firefox' | 'chrome' | 'chromium' | 'default'
MCP_OPENER_BROWSER=firefox
# Explicitly set if Firefox is installed via Flatpak (auto-detected if omitted)
MCP_OPENER_FIREFOX_FLATPAK=true
# Whether to copy files to /tmp for Flatpak access (default: true)
MCP_OPENER_TMP_COPY=true
Claude Code Configuration
Add to your Claude Code MCP settings (.claude/settings.json or global settings):
{
"mcpServers": {
"opener": {
"command": "node",
"args": ["/path/to/mcp-opener/dist/index.js"],
"env": {
"MCP_OPENER_BROWSER": "firefox",
"MCP_OPENER_FIREFOX_FLATPAK": "true",
"MCP_OPENER_TMP_COPY": "true"
}
}
}
}
Or using pnpm:
{
"mcpServers": {
"opener": {
"command": "pnpm",
"args": ["--dir", "/path/to/mcp-opener", "exec", "node", "dist/index.js"],
"env": {
"MCP_OPENER_BROWSER": "firefox"
}
}
}
}
Tools
open_folder
Opens a folder in the system file manager.
Parameters:
path(string, required): Absolute or relative path to the folder
Example:
{
"path": "/home/user/Documents"
}
Behavior by Platform:
- Linux: Uses
xdg-open, falls back tonautilusordolphin - macOS: Uses
open(Finder) - Windows: Uses
explorer
open_file
Opens a file with its default application.
Parameters:
path(string, required): Absolute or relative path to the file
Example:
{
"path": "./report.pdf"
}
Behavior by Platform:
- Linux: Uses
xdg-open - macOS: Uses
open - Windows: Uses
start
open_browser
Opens a URL in the preferred web browser.
Parameters:
url(string, required): URL to open (must start withhttp://,https://, orfile://)
Example:
{
"url": "https://github.com"
}
Behavior:
- Checks configured browser preference (
MCP_OPENER_BROWSER) - On Linux with Firefox:
- Auto-detects Flatpak installation
- For
file://URLs with Flatpak: Copies file to/tmpfor accessibility - Uses
flatpak run org.mozilla.firefoxfor Flatpak - Uses native
firefoxcommand otherwise
- Falls back to system default browser if preferred browser unavailable
Special Handling for Fedora Bluefin / Flatpak Firefox:
When opening local HTML files or other file:// URLs with Flatpak Firefox:
- File is copied to
/tmp/mcp-opener-<random>/ - Browser opens the
/tmpcopy - Cleanup happens after 5 seconds
This works around Flatpak's filesystem sandboxing restrictions.
Platform Support
| Platform | File Manager | Browser Detection | Flatpak Support |
|---|---|---|---|
| Linux | xdg-open, Nautilus, Dolphin | ✅ | ✅ |
| macOS | Finder | ✅ | N/A |
| Windows | Explorer | ✅ | N/A |
Usage Examples
From Claude Code
User: Open my Downloads folder
Agent: [Uses open_folder tool with path: /home/user/Downloads]
User: Open this README file
Agent: [Uses open_file tool with path: ./README.md]
User: Open GitHub in my browser
Agent: [Uses open_browser tool with url: https://github.com]
User: Preview this HTML file
Agent: [Uses open_browser tool with url: file:///home/user/project/index.html]
[On Flatpak Firefox: Copies to /tmp, opens copy, cleans up]
Direct Tool Calls
// Open folder
{
"name": "open_folder",
"arguments": {
"path": "/home/viky/Code"
}
}
// Open file
{
"name": "open_file",
"arguments": {
"path": "./package.json"
}
}
// Open browser
{
"name": "open_browser",
"arguments": {
"url": "https://anthropic.com"
}
}
Error Handling
The server provides detailed error messages:
- File/folder not found:
"File does not exist: /path/to/file" - Invalid URL:
"URL must start with http://, https://, or file://" - Platform not supported:
"Unsupported platform: <platform>" - No file manager found:
"No file manager found (tried xdg-open, nautilus, dolphin)"
Development
# Watch mode for development
pnpm dev
# Build for production
pnpm build
# Test the server manually
echo '{"method":"tools/list"}' | node dist/index.js
Troubleshooting
Flatpak Firefox not detected
Check if Firefox is installed via Flatpak:
flatpak list --app | grep firefox
If installed, you should see org.mozilla.firefox. Set MCP_OPENER_FIREFOX_FLATPAK=true explicitly if auto-detection fails.
Files not opening in Flatpak Firefox
Ensure MCP_OPENER_TMP_COPY=true (default). The server will copy files to /tmp for Flatpak accessibility.
Wrong browser opens
Set MCP_OPENER_BROWSER explicitly:
MCP_OPENER_BROWSER=firefox
# or
MCP_OPENER_BROWSER=chrome
# or
MCP_OPENER_BROWSER=default
License
MIT
Author
3viky
Contributing
Issues and PRs welcome at github.com/3viky/mcp-opener
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。