predictive-maintenance

This MCP server provides AI-assisted predictive maintenance capabilities by exposing industrial diagnostic tools (e.g., vibration analysis, bearing fault detection, ISO 20816-3 related insights) to an LLM via the Model Context Protocol. The server is designed to let a language model orchestrate specialized Python tools to analyze vibration data and generate actionable maintenance guidance. Once running, you can query the server through MCP-compatible clients, asking for diagnostics, reports, or conversational workflows that trigger underlying analysis modules. The goal is to enable natural-language interactions that result in structured diagnostic results, plots, and maintenance recommendations without exposing raw signal processing details to the user.

To use it, start the MCP server and connect a client (like Claude, an LLM, or a local MCP client) to the predictive-maintenance endpoint. The LLM can request tasks such as “generate ISO 20816-3 compliant report for my plant data,” “run bearing fault detection on this vibration sample,” or “summarize envelope analysis results.” The MCP layer translates these requests into calls to the registered Python tools, aggregates results, and returns human-readable summaries, charts, and recommended actions suitable for maintenance teams.

Prerequisites:

• Python 3.11+ installed on your machine
• Access to install Python packages (pip)
• Basic familiarity with running CLI commands

Step-by-step installation:

1. Create and activate a Python virtual environment (optional but recommended): python3.11 -m venv venv source venv/bin/activate # on macOS/Linux venv\Scripts\activate # on Windows

2. Install the MCP server package from PyPI (or from source if provided in repo): pip install predictive-maintenance-mcp

3. Verify installation by listing available MCP servers or running a help command: mcp --help mcp server list

4. Run the MCP server locally (example): mcp run predictive-maintenance

5. Connect an MCP client or LLM to the running server endpoint to begin interactions.

Notes:

• If you prefer using uvx directly, ensure your environment has the uvx launcher configured and accessible in PATH.
• Consult repository docs for any environment-specific configuration files or secrets handling.

Environment variables and configuration tips:

• If the server supports environment-based configuration, you may set variables for data paths, model access, or auth tokens (e.g., PREDICTIVE_MAINT_PATH, MODEL_API_KEY).
• For local air-gapped setups, prefer running the MCP server with a local LLM and keep raw signals on your machine; the MCP layer returns only processed results and summaries.
• If you encounter integration issues, verify that the Python package for the MCP server is installed in the active environment and that the uvx launcher (or equivalent) is properly configured in PATH.
• When extending with new tools, ensure they expose a clear interface (callable Python functions) and are registered in the MCP manifest so the LLM can discover them automatically.