WaveGuardClient

WaveGuardClient is a Python-based MCP server that exposes WaveGuard’s GPU-accelerated anomaly detection via a stateless API call. The server wraps the WaveGuard Python SDK, allowing you to submit any data type—server metrics, transactions, logs, or time series—and receive anomaly scores, confidence levels, and explanations of which features contributed to the alert. It requires only an API key and does not store data between calls, aligning with WaveGuard’s design of being fully stateless. To use, install the WaveGuardClient package, set your API key in the environment, and run the MCP server to listen for requests. The SDK provides a convenient scan API for evaluating training and test data, returning adaptive anomaly scores without any training pipeline.

Prerequisites:

• Python 3.8+ installed on the host
• Network access to install packages from PyPI

Installation steps:

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

2. Install the WaveGuardClient package from PyPI: pip install WaveGuardClient

3. Prepare your environment variable for the API key (recommended): export WAVEGUARD_API_KEY=YOUR_KEY_HERE # Linux/macOS set WAVEGUARD_API_KEY=YOUR_KEY_HERE # Windows

4. Run the MCP server using the provided command (as defined in mcp_config): python -m waveguard

Notes:

• The WaveGuardClient package depends on the requests library; it will be installed automatically via pip.
• If your environment requires additional dependencies (e.g., specific CUDA-enabled GPU drivers), ensure those are configured prior to starting the server.

Tips and considerations:

• Ensure your API key is kept secure and not committed to version control.
• The WaveGuard service is stateless per call; no historical data is stored between requests.
• If you encounter network or authentication errors, verify that WAVEGUARD_API_KEY is correctly set and that the host can reach the WaveGuard API endpoint.
• You can pass various data formats to the scan API (structured data, text, logs). The encoder_type can be adjusted if you’re sending raw text data.
• For local testing, you can run simple Python snippets from the README’s Quickstart to validate basic functionality before integrating into your MCP workflow.