qarai-agent-guard-crewai registers Agent Guard against the CrewAI hook system. It inspects LLM inputs, LLM outputs, tool arguments, and tool results at fixed interception points. It does not change your agent or task definitions.

Installation

pip install qarai-agent-guard qarai-agent-guard-crewai

Quick start

from qarai_agent_guard import (
    AgentGuard,
    Detector,
    default_policy,
)
from qarai_agent_guard_crewai import enable_guard

guard = AgentGuard(
    detectors=[Detector(name="prompt_injection", default_rules="prompt_injection")],
    policy=default_policy(),
)

enable_guard(guard)

By default, enable_guard() registers all four hooks: before_llm_call, after_llm_call, before_tool_call, and after_tool_call.

Documentation

The full documentation has moved to the docs folder. See the CrewAI Integration guide for hooks, parameters, callbacks, and a full workflow example.