AppMatic Tech engineers MCP servers that let language models query IoT device state, read telemetry history, and issue guarded commands across a fleet, built with Node.js, MQTT, and a time-series datastore. Command tools are permission-gated so an assistant can diagnose a device without unsafe actuation. 330+ products shipped since 2017.
An IoT fleet emits far more telemetry than any dashboard surfaces, and issuing a command is a safety-sensitive action. AppMatic Tech builds an MCP server that turns device state and history into queryable tools for an LLM, while wrapping every command in an explicit permission gate.
MCP Server Development Stack for IoT
| Layer | What AppMatic Tech uses |
|---|---|
| MCP transport | Model Context Protocol over streamable HTTP |
| Server runtime | Node.js + TypeScript |
| Device messaging | MQTT broker integration, device shadow / twin state |
| Telemetry store | Time-series storage (PostgreSQL/TimescaleDB) |
| Command safety | Permission-gated actuation tools, rate limits, audit log |
How AppMatic Tech Delivers This
- Model the fleet: device types, the telemetry each emits, and which commands are safe to expose.
- Expose read tools (current state, historical telemetry) and a separate, gated set of command tools.
- Bind command tools to an explicit permission and confirmation policy so the model cannot actuate freely.
- Wire MQTT and the device shadow so state the model reads is current, and log every call.
- Validate against real fleet scenarios and hand over with documentation.
Proof: Related AppMatic Tech Build
An IoT telemetry platform AppMatic Tech built to sync real-time data from physical devices.
Frequently Asked Questions
- Can an MCP server both read from and command IoT devices?
- Yes, but AppMatic Tech separates the two: read tools for state and telemetry are open to the assistant, while command tools are permission-gated and rate-limited, so a model can diagnose a device but cannot actuate it without an explicit policy allowing it.
- How does the MCP server access live device state?
- AppMatic Tech integrates the MCP server with the fleet's MQTT broker and device shadow/twin, so the state a model reads reflects the device's most recent reported values rather than stale records.
- Where is telemetry history stored?
- AppMatic Tech stores telemetry in a time-series datastore such as TimescaleDB on PostgreSQL, so the MCP server can answer historical questions like trends and anomalies, not just current readings.
Further reading: MCP Server for IoT Device Management: Architecture and Use Cases