Sprite MCP Server
Enables management of Sprite VMs through interactive dashboards and terminal interfaces. It allows users to execute commands, create filesystem snapshots, transfer files, and manage VM lifecycles via natural language.
README
Sprite MCP Server
MCP server for managing Sprite VMs with interactive UI dashboards (MCP Apps).
Works with Claude.ai, Claude Desktop, Claude Code, VS Code, and any MCP-compatible client.
Features
- List & manage Sprite VMs - Interactive dashboard with status indicators
- Execute commands - Terminal UI with command history
- Checkpoints - Create, list, restore filesystem snapshots
- File transfer - Push/pull files to/from remote sprites
- Session management - List, attach, kill sessions
Quick Deploy to Your Server
Prerequisites
- Node.js 20+
- A domain pointing to your server (for HTTPS)
- Sprite CLI configured with credentials
1. Clone & Install
git clone https://github.com/davidkiama/sprite-mcp-server.git
cd sprite-mcp-server
npm install
npm run build
2. Configure Environment
cp .env.example .env
# Edit .env with your settings
PORT=3847
SPRITE_BIN=/path/to/sprite # Usually ~/.local/bin/sprite
3. Run with systemd (Recommended)
# Copy service file
sudo cp sprite-mcp.service /etc/systemd/system/
# Edit the service file to set your paths
sudo nano /etc/systemd/system/sprite-mcp.service
# Enable and start
sudo systemctl daemon-reload
sudo systemctl enable sprite-mcp
sudo systemctl start sprite-mcp
# Check status
sudo systemctl status sprite-mcp
4. Set Up Reverse Proxy (Caddy)
# Install Caddy if needed
sudo apt install -y caddy
# Add to /etc/caddy/Caddyfile
echo '
mcp.yourdomain.com {
reverse_proxy localhost:3847
}
' | sudo tee -a /etc/caddy/Caddyfile
# Reload Caddy
sudo systemctl reload caddy
5. Connect to Claude.ai
- Go to claude.ai/settings/integrations
- Add MCP Server:
- Name:
sprite-mcp - URL:
https://mcp.yourdomain.com/sse
- Name:
- Test by asking Claude: "List my sprites"
Local Development
# Run stdio server (for Claude Code/Desktop)
npm start
# Run HTTP/SSE server (for claude.ai)
npm run start:http
# Development mode with hot reload
npm run dev
MCP Tools
| Tool | Description |
|---|---|
list_sprites |
List all Sprite VMs (with interactive dashboard) |
exec_command |
Execute command on a sprite (with terminal UI) |
create_checkpoint |
Create filesystem snapshot |
list_checkpoints |
List available checkpoints |
restore_checkpoint |
Restore to a checkpoint |
get_sprite_url |
Get sprite's public URL |
fetch_file |
Download file from sprite |
push_file |
Upload file to sprite |
list_sessions |
List active sessions |
create_sprite |
Create new sprite VM |
destroy_sprite |
Delete a sprite VM |
Interactive UIs (MCP Apps)
This server uses MCP Apps to provide interactive interfaces:
- Dashboard (
ui://sprite/dashboard) - Visual sprite management - Terminal (
ui://sprite/terminal) - Command execution interface
These render directly in the Claude conversation when using supported clients.
Architecture
┌─────────────────────────────────────────────────────────┐
│ Claude.ai / Desktop │
├─────────────────────────────────────────────────────────┤
│ MCP Protocol │
│ (stdio or HTTP/SSE transport) │
├─────────────────────────────────────────────────────────┤
│ sprite-mcp-server │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────────┐ │
│ │ MCP Tools │ │ UI Resources│ │ Sprite CLI │ │
│ │ (11 tools) │ │ (Dashboard) │ │ Integration │ │
│ └─────────────┘ └─────────────┘ └─────────────────┘ │
├─────────────────────────────────────────────────────────┤
│ Sprite CLI │
│ (sprite list, exec, checkpoint...) │
├─────────────────────────────────────────────────────────┤
│ Sprites API / VMs │
└─────────────────────────────────────────────────────────┘
Troubleshooting
Check if server is running
curl http://localhost:3847/health
# Should return: {"status":"ok","server":"sprite-mcp","version":"1.0.0"}
Check logs
# If using systemd
journalctl -u sprite-mcp -f
# If running directly
npm run start:http 2>&1 | tee mcp.log
Test MCP handshake
echo '{"jsonrpc":"2.0","method":"tools/list","params":{},"id":1}' | npm start
Sprite CLI not found
# Find sprite binary
which sprite
# Set in .env
SPRITE_BIN=$(which sprite)
License
MIT
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。