Crucible: Pytest for AI Agents

Most AI agents today are tested for whether they work.

Very few are tested for how they fail.

That becomes dangerous once agents gain:

• memory

• autonomy

• tool access

• long-running workflows

So we built Crucible:

“Pytest for AI agents.”

An open-source security testing framework for:

• adversarial testing

• behavioral monitoring

• prompt injection evaluation

• agent security scanning

Simple setup:

pip install crucible-security

crucible scan -- target https://your-agent-endpoint.com

Built with:

Python • FastAPI • LangChain • Pydantic

Open-source under Apache 2.0.