# agentsec MCP server

> Scans AI agents for prompt injection, MCP tool poisoning, secret leaks, and other agentic risks using offline Python CLI tools.

- Repository: https://github.com/dunkleschitz/agentsec
- Page: https://hostedmcp.org/servers/dunkleschitz/agentsec
- Transport: —
- Stars: 0
- License: MIT

## FAQ

### What are the requirements to run agentsec?

You only need Python 3.10 or newer. The tools are zero-dependency and use only the standard library.

### Does using these tools guarantee my AI agent is safe?

No, these are heuristic scanners designed to assist human reviewers. A clean result does not guarantee safety.

### Are there any network calls or telemetry involved?

No, the tools are offline and do not make network calls or collect telemetry.

### Can I use the tools as standalone scripts?

Yes, you can clone the repository and run any individual tool as a standalone script without installing the full kit.