LGDiMaggio/predictive-maintenance-mcp

**Give your AI assistant evidence-based vibration diagnostics — machinery fault detection, ISO-cited severity, and diagnostic reports built to support and accelerate expert decision-making.**
An open-source MCP server that turns LLMs into condition monitoring assistants for reliability engineers. Its core design rule: **the server refuses to guess**. No diagnosis is ever inferred from filenames or statistical parameters alone — a fault indication requires matching spectral evidence. Every severity claim cites ISO 20816-3, and the evaluative wording in reports is authored by the server, not improvised by the model. The AI orchestrates the analysis and presents the evidence — detected fault frequencies, matched fault patterns, severity zones — while the final judgment stays with the engineer. Also available as a **Claude Code plugin** with 8 diagnostic skills.
---
See It in Action
Full diagnostic workflow: load signal → spectral analysis → fault detection → severity assessment → report generation
---
Choose Your Path
| You are | Start here |
|---------|------------|
| **Reliability / maintenance engineer** — diagnostics in plain language, no coding | Quick Start below · then Engineer's guide for your own data, manuals and workflows |
| **AI / MCP developer** — run, integrate, and extend the server | Developer's Quickstart · Installation |
| **Researcher / evaluator** — how the numbers are measured | Benchmark Methodology · Benchmark below |
---
What Can It Do?
**Point the AI at a vibration signal → get the evidence behind the fault — detected frequencies, matched fault patterns, ISO-cited severity — to support your call.**
| You say | The AI does |
|---------|-------------|
| *"Is this bearing healthy?"* | Loads the signal, runs spectral analysis, surfaces matching fault-frequency evidence, cites the ISO 20816-3 severity zone |
| *"Generate a full diagnostic report"* | Produces an interactive HTML report with charts, fault markers, and server-authored severity wording |
| *"Extract specs from test_pump_manual.pdf and diagnose the signal"* | Reads the equipment manual, looks up the bearing model, calculates expected fault frequencies, flags which ones the signal actually shows |
| *"Train an anomaly detector on my healthy baselines, then flag anomalies"* | Trains a model on your normal data, scores new signals, flags outliers for your review |
| *"What changed on pump P-101 since the baseline?"* | Reads the asset's recorded history, compares the latest acquisitions with the declared reference, and reports each indicator's change with the criterion it applied |
The AI doesn't guess: it calls **41 specialized MCP endpoints** (38 tools + 3 prompts) running locally on your machine. Every signal is referenced by a single signal_id handle from load to report. Your data never leaves your infrastructure.
Full endpoint reference, grouped by category: **Tool Catalog**.
---
Quick Start
One route, from a fresh clone to a first diagnosis in Claude Desktop. You need Git (or the repository ZIP from GitHub, extracted), uv (it fetches Python 3.11 by itself) and Claude Desktop.
**1. Get the repository.** The 20 sample signals ship with it, not with the PyPI package: real bearing measurements from the MathWorks rolling-element bearing dataset, healthy baselines plus inner and outer race faults, each with its companion metadata (data/README.md).
**2. Install and connect.** On Windows, one script creates the environment, installs the server and writes the Claude Desktop entry (clones inside OneDrive or other synced folders are handled):
macOS
Add this entry to ~/Library/Application Support/Claude/claude_desktop_config.json, with the absolute path of your clone in place of /path/to/predictive-maintenance-mcp. If a predictive-maintenance entry is already there, replace it:
PDM_PROJECT_DIR tells the server where the sample signals and the reports live. Keep every path absolute: Claude Desktop launches servers with a minimal PATH.