yzhao062/pyod

A Python library for anomaly detection across tabular, time series, graph, text, image, and audio data. 60+ detectors, benchmark-backed ADEngine orchestration, and an agentic workflow for AI agents.

What it solves

PyOD is a comprehensive toolkit for anomaly detection (outlier detection) across various data types. It simplifies the process of identifying abnormal patterns in data, providing a unified API for a vast array of algorithms, and automating the selection and assessment of the best detector for a given dataset.

How it works

PyOD provides three levels of interaction:

  1. Classic API: A standard fit/predict interface for users who already know which specific algorithm (e.g., Isolation Forest) they want to use.
  2. ADEngine: An orchestration core that automatically chooses, compares, and assesses multiple detectors to find the most effective one.
  3. Agentic Investigation: An AI-driven layer where users can interact with the library via natural language. This is powered by the od-expert skill (for Claude Code/Codex) and an MCP server that exposes tools for knowledge queries, planning, and detection to LLM agents.

Who it’s for

  • Data Scientists and ML Engineers: Who need a scalable, benchmark-backed library to detect outliers in tabular, time series, graph, text, image, or audio data.
  • AI Agent Developers: Who want to integrate professional anomaly detection capabilities into LLM-based agents via MCP or specific skills.
  • Researchers: Who require a standardized framework to benchmark different outlier detection algorithms.

Highlights

  • Multi-Modal Support: Includes 61 detectors covering tabular, time series, graph, text, image, and audio data.
  • Agent-Ready: Integrates with Claude Code, Codex, and MCP-compatible LLMs to turn natural language requests into detection workflows.
  • Full Lifecycle Management: Handles everything from raw data profiling to explained anomalies and next-step guidance.
  • High Performance: Built on SUOD for parallel training and utilizes Numba JIT for per-model speedups.
  • Widespread Adoption: Over 46 million downloads and used by organizations like the European Space Agency and Walmart.

Related

  • Project
    Albertsr/Anomaly-Detection

    A code-first educational collection of handwritten Python implementations for various tabular anomaly detection methods, including PCA reconstruction, RobustPCC, and PU learning.

  • Project
    unit8co/darts

    A Python library for user-friendly time series forecasting and anomaly detection, offering a unified API for models ranging from classic statistical methods to deep neural networks.

  • Project
    kkirchheim/pytorch-ood

    pytorch‑ood is a genuine Python library for out‑of‑distribution detection built on PyTorch. It bundles >30 detectors, loss functions, pretrained models, datasets and utilities, integrates with pytorch‑lightning, and offers a simple API plus benchmark helpers. Install via `pip install pytorch-ood` and start using detectors like `EnergyBased` with just a few lines of code.

  • Project
    vc1492a/PyNomaly

    A Python implementation of Local Outlier Probabilities (LoOP) that provides interpretable outlier scores as probabilities between 0 and 1.

  • Project
    K-Dense-AI/agentic-data-scientist

    An adaptive multi-agent framework that automates complex data science workflows by separating planning from execution and utilizing a self-correcting, iterative validation loop.