PentrAI is an automated penetration-testing tool. A company can point it at its own infrastructure to run security assessments and surface the exposures worth fixing, instead of commissioning a manual engagement for every check.
An LLM agent drives a catalog of security tools through a single server over the
Model Context Protocol (MCP): a Flask backend (pentrai_server.py) wraps each
tool as an HTTP route, and an MCP client (pentrai_mcp.py) presents those tools
to the agent.
python3 -m venv pentrai-env
source pentrai-env/bin/activate
pip install -r requirements.txt
python3 pentrai_server.py # serves on 127.0.0.1:8888 by defaultPoint an MCP-capable client at the server using pentrai-mcp.json as a template
(set the script path and the server address). Override the bind address with the
PENTRAI_HOST environment variable.
The server shells out to external security tools; install the ones you need from
their official sources, alongside the Python libraries in requirements.txt.
Early stage. This README is intentionally short; fuller documentation will follow as the product develops.
MIT. See LICENSE.