Open-source MCP servers bridging AI to physical hardware.
AI can control Jira, GitHub, Slack — but what about your oscilloscope, Arduino, or PLC?
hardware-mcp is building the protocol layer that connects AI agents to the physical world.
| Server | Hardware | Status |
|---|---|---|
| arduino-mcp-server | Arduino / arduino-cli | ✅ Stable |
| scpi-mcp-server | Oscilloscopes, PSUs, signal generators (Rigol, Keysight, Tektronix) | 🚧 Coming soon |
| jtag-mcp-server | Embedded targets via JTAG/SWD (OpenOCD, J-Link, CMSIS-DAP) | 🚧 Coming soon |
| canbus-mcp-server | CAN bus — automotive and industrial networks | 🚧 Coming soon |
| modbus-mcp-server | Industrial PLCs / Modbus RTU & TCP | 🚧 Coming soon |
Model Context Protocol (MCP) is an open standard by Anthropic that lets AI models interact with external tools and services through a structured interface.
hardware-mcp extends this to the physical world — giving AI agents the ability to read sensors, control actuators, compile and upload firmware, monitor serial output, and run safety checks before touching real hardware.
- Safety first — every server includes electrical safety preflight checks
- Open source — MIT licensed, community driven
- Protocol-level thinking — we're building a standard, not just wrappers
- Research-backed — we publish benchmarks and papers on AI-hardware control
- ⭐ Star a server repo
- 🐛 File issues or feature requests
- 🔧 Contribute a new hardware server
- 📄 Read our research (coming soon)
Built by engineers who work with hardware every day.