simple-rail-mcp
Enables controlling linear rails via natural language commands through Claude AI in Cursor, supporting HTTP, serial, and G-code controllers with configurable positions.
README
Simple Rail MCP Server
MCP (Model Context Protocol) server for controlling linear rails via natural language through Claude AI in Cursor.
What This Does
Control your linear rail with natural language:
You: "Move the rail to position 500mm"
Claude: ✅ Rail moved to 500mm
You: "Move rail to opentrons station"
Claude: ✅ Rail at opentrons station
Supported Controllers
- ✅ UW PICO 5.09 - HTTP API with G-code forwarding (pre-configured)
- ✅ Any HTTP API that accepts commands
- ✅ Serial port controllers
- ✅ Python script wrappers
- ✅ G-code senders
Quick Start
1. Install
npm install
2. Configure
Edit index.js lines 32-37:
// For UW PICO controller:
this.RAIL_IP = "192.168.1.101"; // Your controller IP
this.RAIL_PORT = "80";
this.RAIL_ENDPOINT = "/command";
this.COMMAND_TEMPLATE = "G0 X{position} F1000";
Set your positions (lines 41-47):
this.POSITIONS = {
"pickup_station": 0,
"opentrons": 500,
"plate_reader": 750,
"storage": 1000,
"home": 0
};
3. Add to Cursor
Edit .cursor/mcp.json:
{
"mcpServers": {
"rail": {
"command": "node",
"args": ["/absolute/path/to/simple-rail-mcp/index.js"],
"env": {
"RAIL_IP": "192.168.1.101"
}
}
}
}
4. Restart Cursor
5. Use!
"Move the rail to opentrons"
"Move rail to 500mm"
"Move to home position"
Configuration Options
For UW PICO 5.09 Controller
Your controller has an HTTP API that forwards G-code via UART.
Test with curl:
curl -X POST http://192.168.1.101/command \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "cmd=G0 X500 F1000"
Config:
this.RAIL_IP = "192.168.1.101";
this.RAIL_PORT = "80";
this.RAIL_ENDPOINT = "/command";
this.COMMAND_TEMPLATE = "G0 X{position} F1000";
For Other HTTP APIs
If your controller uses different endpoints/formats:
this.RAIL_IP = "192.168.1.100";
this.RAIL_ENDPOINT = "/api/move";
this.COMMAND_TEMPLATE = '{"position": {position}}'; // JSON format
Position Configuration
Edit the POSITIONS object to match your physical setup:
this.POSITIONS = {
"station_a": 0, // First position (mm)
"station_b": 250, // Second position (mm)
"station_c": 500, // Third position (mm)
"home": 0 // Home position
};
Timing Configuration
Adjust wait time after movement (line 54):
this.MOVE_WAIT_TIME = 5; // Wait 5 seconds after sending command
If your rail is faster or slower, adjust accordingly.
Environment Variables
You can override config with environment variables:
{
"mcpServers": {
"rail": {
"command": "node",
"args": ["/path/to/index.js"],
"env": {
"RAIL_IP": "192.168.1.101",
"RAIL_PORT": "80"
}
}
}
}
API Reference
Tool: move_rail_and_wait
Move linear rail to a position and wait for completion.
Parameters:
position(string) - Named position fromPOSITIONSconfig- Example:
"opentrons","pickup_station","home"
- Example:
position_number(number) - Numeric position in mm- Example:
500,750.5
- Example:
wait_seconds(number) - Override default wait time- Default:
MOVE_WAIT_TIMEfrom config
- Default:
Examples:
// Via Claude in natural language:
"Move rail to opentrons"
"Move rail to 500mm"
"Move to home position"
// Direct tool call (from code):
move_rail_and_wait({ position: "opentrons" })
move_rail_and_wait({ position_number: 500 })
move_rail_and_wait({ position: "storage", wait_seconds: 10 })
G-code Reference
Common G-code commands for linear rails:
| Command | Description |
|---|---|
G0 X500 F1000 |
Move to 500mm at 1000mm/min (fast) |
G1 X500 F500 |
Move to 500mm at 500mm/min (controlled) |
G28 X |
Home X axis |
G90 |
Set absolute positioning mode |
G91 |
Set relative positioning mode |
G92 X0 |
Set current position as zero |
M114 |
Query current position |
M400 |
Wait for all moves to complete |
Troubleshooting
Cannot connect to controller
-
Check IP:
ping 192.168.1.101 -
Test endpoint:
curl http://192.168.1.101/command -
Check firewall - Allow Node.js/Cursor through firewall
Rail doesn't move
-
Test G-code manually:
curl -X POST http://192.168.1.101/command -d "cmd=G28 X" # Home first curl -X POST http://192.168.1.101/command -d "cmd=G0 X500 F1000" -
Check positioning mode:
curl -X POST http://192.168.1.101/command -d "cmd=G90" # Absolute mode -
Verify position is in range - Check your rail's physical limits
Wrong position
-
Check units - Make sure G21 (mm) is set:
curl -X POST http://192.168.1.101/command -d "cmd=G21" -
Calibrate positions - Measure actual distances and update
POSITIONS -
Home the rail first:
curl -X POST http://192.168.1.101/command -d "cmd=G28 X"
MCP server not loading in Cursor
-
Test manually:
node index.js -
Check Node version:
node --version # Should be ≥18.0.0 -
Check path in
.cursor/mcp.json- Use absolute path -
Restart Cursor completely
Advanced: Custom Controller Support
To add support for other controller types, modify the sendHttpCommand method:
async sendHttpCommand(position) {
const gcode = this.COMMAND_TEMPLATE.replace('{position}', position);
const url = `http://${this.RAIL_IP}:${this.RAIL_PORT}${this.RAIL_ENDPOINT}`;
// Customize request format here
const response = await fetch(url, {
method: 'POST',
headers: {
'Content-Type': 'application/json' // Change as needed
},
body: JSON.stringify({ command: gcode }) // Change format
});
// Customize response parsing here
const data = await response.json();
return data;
}
Integration Example
Use with other MCP servers for full automation:
// Example: Pick up plate and move to Opentrons
"Move rail to pickup station" // This MCP server
"Run VLA pickup script" // VLA MCP server
"Move rail to opentrons" // This MCP server
"Upload purple_mixing.py" // Opentrons MCP server
"Start the protocol" // Opentrons MCP server
Claude AI automatically orchestrates all systems!
Requirements
- Node.js ≥18.0.0
- Cursor IDE with MCP support
- Linear rail with HTTP/serial/script control
License
MIT
Contributing
Issues and pull requests welcome!
Credits
- Model Context Protocol - Anthropic
- Claude AI - Natural language orchestration
- Cursor IDE - Development environment
Questions? Open an issue or check the troubleshooting guide.
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。