arduino-claude-mcp
MCP server for Arduino IDE 2.0 sketches. It exposes a small REST API plus an MCP stdio bridge so agents can read and write the main .ino file, list sources, and optionally compile with arduino-cli.
README
arduino-claude-mcp
MCP server for Arduino IDE 2.0 sketches. It exposes a small REST API plus an MCP stdio bridge so agents can read and write the main .ino file, list sources, and optionally compile with arduino-cli.
Features
- Validate an Arduino sketch folder
- Read and write source files (defaults to main
.ino) - List .ino/.h/.cpp/.c files
- Compile with
arduino-cli(optional) - REST + MCP stdio transport
Requirements
- An Arduino sketch folder (Arduino IDE 2.0 format)
- Node.js 18+
- Optional:
arduino-clifor builds
Installation
Global install:
npm install -g arduino-claude-mcp
Local dev:
npm install
npm run build
npm run build:mcp
Container image:
docker pull eoinedge/arduino-mcp:latest
Usage
Run the REST server
arduino-claude-mcp
Defaults to port 3080. Override with:
$env:PORT=3081
arduino-claude-mcp
Run with Docker (Pi/OpenClaw)
docker run --rm --network host \
-e PORT=3080 \
-e ARDUINO_FQBN=arduino:mbed_nano:nano33ble \
-v /home/pi/pi-openclaw-mcp-stack/workspace/Arduino:/workspace \
eoinedge/arduino-mcp:latest
Run the MCP stdio server
node build/mcp.js
If your REST server is not on the default port, set one of:
ARDUINO_API_URL(full URL, for examplehttp://localhost:3081)ARDUINO_API_PORT(port only, for example3081)
MCP client config
{
"mcpServers": {
"arduino-mcp": {
"command": "node",
"args": ["/absolute/path/to/build/mcp.js"]
}
}
}
Tutorial (10 minutes)
- Create a sketch folder whose main
.inomatches the folder name. - Validate it:
POST /validate- Body:
{ "projectRoot": "/path/to/MySketch" }
- Read the main sketch:
POST /read_source- Body:
{ "projectRoot": "/path/to/MySketch" }
- Write a small change:
POST /write_source- Body:
{ "projectRoot": "/path/to/MySketch", "content": "// your code" }
- Build (optional):
- Set
ARDUINO_FQBN(examplearduino:avr:uno) POST /build- Body:
{ "projectRoot": "/path/to/MySketch" }
- Set
Environment variables
PORT: REST server port (default3080)ARDUINO_API_URL: REST base URL for the MCP bridgeARDUINO_API_PORT: REST port for the MCP bridgeARDUINO_CLI: arduino-cli executable (defaultarduino-cli)ARDUINO_FQBN: fully qualified board name for compileARDUINO_BUILD_TIMEOUT_MS: build timeout in ms (default120000)
REST API
Health
GET /health->{ status: "ok" }
Validate
POST /validatebody:{ projectRoot: string }- Returns
{ valid, inoPath, projectRoot }
List sources
POST /list_sourcesbody:{ projectRoot: string }- Returns
{ files: string[] }
Read source
POST /read_sourcebody:{ projectRoot: string, relativePath?: string }- Returns
{ path, content }
Write source
POST /write_sourcebody:{ projectRoot: string, relativePath?: string, content: string }- Returns
{ success, path, bytes }
Append source
POST /append_sourcebody:{ projectRoot: string, relativePath?: string, content: string }- Returns
{ success, path, bytes }
Build
POST /buildbody:{ projectRoot: string, timeoutMs?: number }- Returns
{ success, exitCode, stdout, stderr } - Requires
arduino-cliandARDUINO_FQBN
Clawdbot / Moltbot compatibility
This repo ships a skill at skills/arduino-mcp/SKILL.md.
Enable it in Moltbot:
{
"skills": {
"load": {
"extraDirs": ["~/.clawdbot/skills"],
"watch": true,
"watchDebounceMs": 250
},
"entries": {
"arduino-mcp": {
"enabled": true,
"env": {}
}
}
}
}
Example prompts
- "Open this Arduino sketch and add a blinking LED on pin 13."
- "List all source files and explain what each does."
- "Append a serial debug line and recompile for an Uno board."
Testing
npm test
Tests use a temporary sketch folder and do not require arduino-cli.
Publishing
- Update
package.jsonversion - Build:
npm run build npm run build:mcp - Publish:
npm publish --access public - Tag and push to trigger
.github/workflows/publish.yml
Docker publishing
Manual push:
docker login
docker buildx build --platform linux/amd64,linux/arm64 \
-t eoinedge/arduino-mcp:latest \
-t eoinedge/arduino-mcp:<version> \
--push .
GitHub Actions:
docker-publish.ymlpushes Docker images onmainand version tags.- Required repo secrets:
DOCKERHUB_USERNAMEDOCKERHUB_TOKEN
Project structure
src/index.tsREST APIsrc/mcp.tsMCP stdio bridgeskills/arduino-mcp/SKILL.mdagent skillserver.jsonMCP registry metadata
Contributing
PRs welcome. Please keep changes small and include tests where possible.
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。
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。
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。
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。