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.
PyOD is a comprehensive Python library for anomaly detection that includes AI agent integration through the od-expert skill and MCP server. It offers 61+ detectors across tabular, time series, graph, text, image, and audio data with a unified API. The library enables AI agents to drive anomaly detection investigations through natural language via the od-expert skill (for Claude Code and Codex) or structured MCP tools (for any MCP-compatible LLM), while maintaining classic fit/predict APIs for direct Python use. With 46+ million downloads and benchmark-backed performance, it supports both automated detector selection through ADEngine and manual detector choice.
Key Features
Use Cases
- 01Enabling AI agents to investigate anomalies in datasets through conversational multi-turn dialogues
- 02Detecting fraudulent transactions, network intrusions, or unusual system behavior in tabular data
- 03Identifying anomalies in time-series telemetry from spacecraft, industrial sensors, or financial markets
- 04Finding outlier nodes in social networks, citation graphs, or cybersecurity attack graphs
- 05Detecting unusual text documents, spam, or anomalous content via NLP embeddings
- 06Automated model selection and hyperparameter tuning for anomaly detection pipelines using ADEngine
Related Skills
View moresuperpowers
An agentic skills framework & software development methodology that works.
skills
Skills for Real Engineers. Straight from my .agents directory.
skills
Public repository for Agent Skills
ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
pyod — FAQ
What is PyOD and what does the od-expert skill do?+
PyOD is a Python anomaly detection library with 61+ algorithms. The od-expert skill allows Claude Code and Codex agents to drive multi-turn anomaly detection investigations through natural language, automatically selecting detectors, analyzing data, and explaining findings.
How do I install the od-expert skill for Claude Code?+
First install PyOD with 'pip install pyod', then run 'pyod install skill' for user-global installation in Claude Code (~/.claude/skills/). For project-local Codex installation, use 'pyod install skill --project'.
Which AI clients work with PyOD's agent features?+
The od-expert skill works with Claude Code and Codex. The MCP server (via 'pyod mcp serve') works with any MCP-compatible LLM client that supports the Model Context Protocol.
Do I need API keys or additional dependencies?+
The core library requires only 'pip install pyod'. The MCP server requires 'pip install pyod[mcp]'. Optional encoders like OpenAI embeddings require API keys, but most detectors and local encoders (sentence-transformers, HuggingFace) work without external APIs.
Is PyOD free to use?+
Yes, PyOD is open source and free to use. It has been downloaded 46+ million times and is used in both academic research and commercial products.
Can I use PyOD without AI agents?+
Absolutely. PyOD's classic fit/predict API works as a standalone Python library with scikit-learn compatibility. The agent features (od-expert skill and MCP server) are optional Layer 3 capabilities built on top of the core library.
How do I install pyod?+
Open the source repository on GitHub and follow its README. pyod is a skill — MCP Agents Market links you directly to the official repo.
Is pyod free?+
pyod is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.