# Pantheon Security — MEDUSA

> MEDUSA is a free, open-source (AGPL-3.0) AI-first security scanner with 79 scanners and 40,000+ detection rules for LLM agents, MCP servers, RAG pipelines, prompt injection, and traditional code vulnerabilities. It works out of the box — install with `pip install medusa-security` and run `medusa scan .` — no external tools, API keys, or configuration required. Current release: v2026.7.0.

MEDUSA detects prompt injection and jailbreaks, MCP tool poisoning, RAG/dataset poisoning, agent-protocol attacks (MCP, UCP, AP2, ACP), AI supply-chain attacks (including poisoned AI editor configs such as .cursorrules, mcp.json, and .claude/ hooks), leaked secrets, and 200 known CVEs across 7 package ecosystems. All scanning runs locally; no code or results are transmitted.

## Pages

- [Home](https://pantheonsecurity.io/): product overview, features, coverage, pricing, FAQ
- [Documentation](https://pantheonsecurity.io/docs): quick start, installation, scanner reference
- [Security Policy](https://pantheonsecurity.io/security): vulnerability reporting and disclosure process
- [Privacy Policy](https://pantheonsecurity.io/privacy)
- [Terms of Service](https://pantheonsecurity.io/terms)

## Project

- [GitHub repository](https://github.com/Pantheon-Security/medusa): source code, issues, releases
- [PyPI package](https://pypi.org/project/medusa-security/): `pip install medusa-security`
- [Changelog](https://github.com/Pantheon-Security/medusa/blob/main/CHANGELOG.md)