# RogueProof MCP server

> Test and stress-test autonomous AI agents by simulating attacks and evaluating them against different security policies and budget constraints.

- Repository: https://github.com/ivanmolto/rogueproof
- Page: https://hostedmcp.org/servers/ivanmolto/rogueproof
- Transport: —
- Stars: 0
- License: Apache-2.0

## FAQ

### Are the harmful actions performed by the agent real?

No, harmful actions are simulated. Secrets are fake, the attacker's sink is a mock, and the agent's commands run in a container with no network access.

### How does RogueProof ensure the evidence is not faked?

The system uses real models, real Otari requests, real costs, and real refusals to ensure all AI decisions are authentic.