What is Predictive Maintenance MCP Server?

An open-source AI agent that lets you ask an AI assistant about your machine's vibration data and get real, evidence-based maintenance answers, running entirely on your computer. It provides 41 MCP endpoints for vibration analysis, bearing fault detection, ISO 20816 severity assessment, and ML anomaly detection.

What kinds of machine faults can this AI agent detect?

Outer race, inner race, and rolling element bearing defects; gear mesh issues; general unbalance and misalignment. The system uses frequency analysis, envelope detection, and trained anomaly models.

Build Your Predictive Maintenance AI Agent — Open-Source MCP Server & Claude Plugin

WHY THIS EXISTS

Tools turn guesses into evidence

Same AI, different outcomes. Without diagnostic tools an LLM can only speculate. With them, it runs a real signal-processing pipeline and answers with measured values.

✕ PLAIN LLM, NO TOOLS

"It might be a bearing fault… maybe check the alignment? I'd suggest monitoring it."

· Text-only: never touched your data · No measurements, no frequencies · Not repeatable across operators

✓ WITH MCP DIAGNOSTIC TOOLS

"Outer-race bearing wear detected at 89.4 Hz (expected 89.8 Hz). Vibration: 4.2 mm/s, alert zone. Recommend inspection within 24 h."

· Real sensor data, processed locally · Measured frequencies and ISO severity · Consistent, auditable results

Raw waveforms stay on your machine: only computed summaries (severity scores, fault classes) reach the model.

GET STARTED

From install to first diagnosis in 5 minutes

1 Install pip install predictive-maintenance-mcp Then connect to Claude Desktop, VS Code, or any MCP client. Full setup guide →

2 Ask "Load OuterRaceFault_1.csv and check if the bearing is healthy." Sample bearing data is included in the repository. Clone it and try a diagnosis with no sensors connected.

3 Act Get a diagnosis with measured evidence, an ISO 20816 risk class, confidence level, and a shareable HTML report for your team.

CAPABILITIES

Grouped by what you get done

DIAGNOSE Find the fault Decompose signals into frequency components, detect impact patterns from early bearing or gear wear, and match measured peaks against expected fault signatures.

FFT + peak detection Envelope analysis Bearing & gear signatures Manual-aware (RAG)

ASSESS RISK Know how urgent it is Automatic severity zones per ISO 20816 and ML anomaly scores trained on your healthy baselines, so the team knows what needs attention now and what can wait.

ISO 20816 zones Anomaly detection RUL estimation

DECIDE Act with confidence Evidence-based recommendations with confidence levels, plus HTML and Word reports designed for engineers, operators, and management. Every claim cites its source data.

Recommendations Confidence levels HTML / Word reports

EXTEND Build on it Plugin architecture for custom signal processors and report formats. Ingest CSV, WAV, MAT, NPY, and Parquet. Works with any MCP-compatible client, including air-gapped Ollama.

41 MCP endpoints Custom processors Multi-format ingestion Ollama air-gapped

UNDER THE HOOD

Your raw data never leaves your machine

You

natural language

→

AI client

Claude · GPT · Ollama

→

PM-MCP server

41 MCP endpoints

→

Your data, local

signals · manuals · models

→

Computed evidence

severity · faults · reports

Privacy-first Raw waveforms are processed entirely locally. Only computed results flow to the LLM.

Works with any AI Claude, ChatGPT, Copilot Studio, or Ollama for fully air-gapped deployments.

Six-stage pipeline Ingest → process → detect → assess → predict → recommend. Each stage independent and testable.

CLAUDE CODE PLUGIN

Repeatable workflows inside your editor

For developers and power users. The MCP server answers questions; the plugin automates whole diagnostic pipelines.

8 Domain skills Bearing diagnosis, gear analysis, screening, prognostics, reports, anomaly detection, signal management, docs search.

2 Workflow agents diagnostic-pipeline runs signal → report end-to-end; signal-explorer compares multiple signals.

3 Slash commands /pm-diagnose · /pm-screen · /pm-report

$ /plugin marketplace add LGDiMaggio/predictive-maintenance-mcp

READY TO TRY

Real bearing data included

20 real bearing vibration signals are included in the repository: 3 healthy baselines plus inner-race and outer-race fault signals from the MathWorks rolling-element bearing dataset. Clone the repo and diagnose your first fault without connecting a single sensor.

"Load OuterRaceFault_1.csv and diagnose the bearing."

Dataset: MathWorks RollingElementBearingFaultDiagnosis-Data (CC BY-NC-SA 4.0). Not included in the PyPI package.

Healthy baseline 3 signals · ZONE A

Inner-race fault BPFI signature

Outer-race fault BPFO signature

QUESTIONS

Frequently asked

What is this project, in one sentence?+

An open-source tool that lets you ask an AI assistant about your machine's vibration data and get real, evidence-based maintenance answers, running entirely on your computer.

Does my data leave my machine?+

No. Raw vibration files are processed locally. Only computed summaries (severity scores, fault classifications) are sent to the AI for interpretation. Use Ollama for a fully air-gapped setup.

Which AI models are supported?+

Any MCP-compatible client: Claude Desktop, VS Code with Copilot, ChatGPT (via MCP bridge), Microsoft Copilot Studio, or local models with Ollama.

Do I need signal-processing knowledge?+

No. Ask questions in plain language like "Is this bearing healthy?" The AI handles the analysis pipeline and explains results in everyday terms.

What kinds of faults can it detect?+

Outer race, inner race, and rolling element bearing defects; gear mesh issues; general unbalance and misalignment, via frequency analysis, envelope detection, and trained anomaly models.

Is this a replacement for professional vibration analysis?+

No. It's a screening and decision-support tool that helps prioritize and communicate. Critical decisions should always be validated by qualified reliability engineers.

Can I add my own analysis tools?+

Yes. The plugin architecture supports custom signal processors, diagnostic logic, and report formats. See the Developer Quick Start.

PM-MCP is a screening and decision-support tool. Critical maintenance decisions should always be validated by qualified reliability engineers; it does not replace certified vibration analysis.

Ask your machines what's wrong. With evidence, not guesses.