An MCP server that gives AI assistants evidence-based vibration diagnostics and ISO-cited machinery fault reports.
Turn your AI desktop assistant into a condition monitoring aid for pumps, motors, and bearing diagnostics.
This tool is a Python-based Model Context Protocol (MCP) server that connects LLM clients like Claude Desktop to local vibration analysis engines. It suits facility maintenance leads and engineers in hospitality infrastructure who want AI-assisted equipment fault detection anchored strictly in ISO standards without sending sensor data to third-party clouds.
It exposes 42 specialized MCP endpoints that execute envelope analysis, calculate bearing fault frequencies, and assess ISO 20816-3 severity zones. Instead of letting LLMs guess root causes from filenames or statistical summaries alone, the server forces the AI to present matching spectral evidence and outputs interactive HTML diagnostic reports.
It suits technical facility managers, reliability engineers, and operations teams managing pumps, HVAC systems, or guest-service machinery who already use desktop AI assistants. It is not suitable for non-technical staff expecting a turn-key web UI, as it requires configuring local Python environments and preparing single-column CSV signal files.
Requires spectral line evidence above noise floors before confirming fault indications rather than letting the LLM guess.
Evaluates vibration signals directly against ISO standard velocity thresholds to assign machine severity zones.
Generates self-contained diagnostic report files featuring annotated spectral charts and server-authored evaluative text.
Runs completely on your local machine so raw vibration data and PDF equipment manuals never leave your network.
Trains statistical anomaly detectors on healthy operational baselines to flag potential outliers in newly acquired signals.
Free under the MIT license. The only cost is running it on your own local hardware and paying for any underlying API usage required by your chosen MCP client.
873 open remote AI-training roles, updated November 15, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
No, the MCP server runs entirely on your local machine and processes all signal files locally.
Yes, you can instruct your AI client to read PDF manuals, extract bearing model specs, calculate expected fault lines, and test them in the server.
See Zekai first in Google