Arm's official MCP server for porting and tuning software on Arm — semantic docs search, codebase migration scanning, container architecture checks and LLVM-MCA assembly analysis.

tinymcpMCP Server
Summary
Golioth's experimental MCP server that lets an LLM drive real embedded devices — device RPCs become MCP tools, with schemas stored in LightDB State and calls routed through the Golioth cloud.
Features
- Exposes device RPCs to an LLM as MCP tools
- Tool schemas defined per device via Golioth LightDB State
- Reaches remote and battery-powered devices through the Golioth cloud
- Go server with Zephyr RTOS device firmware
- Apache-2.0 licensed
Installation
Set up this MCP server in your favorite AI agent — copy a ready-made configuration below.
Description
tinymcp is Golioth's experimental bridge between Model Context Protocol clients and physical embedded hardware. Instead of a server that reads documentation about devices, this one actually operates them: an LLM can turn on an LED, read a sensor, or trigger a reboot on a microcontroller sitting on another continent.
How it works. The MCP server runs locally and talks to hardware through the Golioth device management cloud rather than over a direct connection, which is what makes remote and sleepy battery-powered devices reachable at all. Two Golioth services do the work:
- Remote Procedure Calls (RPCs) — each RPC a device registers becomes an MCP tool the client can call.
- LightDB State — holds the tool schemas that describe those RPCs, so the set of available tools is defined by the device itself rather than hardcoded in the server.
Running it. The server is written in Go (go build -o tinymcp ./server) and is configured with three environment variables: TINYMCP_PROJECT, TINYMCP_DEVICE and TINYMCP_API_KEY. Device-side firmware targets Zephyr RTOS and is fetched with west. The project is Apache-2.0 licensed.
Worth knowing. The maintainers label this experimental. Giving a language model a callable path to physical actuators deserves a scoped API key and a device you do not mind rebooting.
Related MCP Servers
Official AntV MCP server that turns a description of your data into a rendered chart — 26 visualization types from bar and line to sankey, mind-map, fishbone and geographic maps.
Context-compression MCP server that shrinks tool outputs, logs, files and RAG chunks before they reach the model, cutting 60-95% of tokens on JSON payloads.
Persistent memory for coding agents as a single Go binary — SQLite and FTS5 behind MCP, an HTTP API, a CLI and a TUI, with no Node, Python or Docker required.