Hackagent

HackAgent is an open source red team toolkit for testing AI agents against prompt injection, jailbreaking and goal hijacking. It is installed as a Python package with pip, and a developer points it at an agent endpoint, chooses an attack configuration and runs it from a few lines of code. The toolkit sends adversarial goals to the target agent and uses a judge model to decide whether the agent's safety guardrails were bypassed. Attack types shown on the site include advprefix and pair, and the judge configuration references the HarmBench evaluation type. It supports agents built with Google ADK, the OpenAI SDK, LiteLLM and Ollama, and it can also test Claude Code and Codex CLI. A companion web dashboard lets a team register agents, review attack runs and their jailbreak success rates, and read security reports that break findings down by severity, with counts of tests and vulnerabilities per agent. It is built for developers and security teams who run their own agents and want to check them before deployment. It tests systems the user controls and supplies the endpoint for. Documentation and source code are linked from the home page, and the dashboard requires a login. The package is open source and free to install. The site does not publish a pricing page for the hosted dashboard.